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(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         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(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         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 /* Validate changes from /proc interface. */
 142 static int proc_tcp_default_init_rwnd(ctl_table *ctl, int write,
 143                                       void __user *buffer,
 144                                       size_t *lenp, loff_t *ppos)
 145 {
 146         int old_value = *(int *)ctl->data;
 147         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 148         int new_value = *(int *)ctl->data;
 149
 150         if (write && ret == 0 && (new_value < 3 || new_value > 100))
 151                 *(int *)ctl->data = old_value;
 152
 153         return ret;
 154 }
 155
 156 static int proc_tcp_congestion_control(ctl_table *ctl, int write,
 157                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 158 {
 159         char val[TCP_CA_NAME_MAX];
 160         ctl_table tbl = {
 161                 .data = val,
 162                 .maxlen = TCP_CA_NAME_MAX,
 163         };
 164         int ret;
 165
 166         tcp_get_default_congestion_control(val);
 167
 168         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 169         if (write && ret == 0)
 170                 ret = tcp_set_default_congestion_control(val);
 171         return ret;
 172 }
 173
 174 static int proc_tcp_available_congestion_control(ctl_table *ctl,
 175                                                  int write,
 176                                                  void __user *buffer, size_t *lenp,
 177                                                  loff_t *ppos)
 178 {
 179         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 180         int ret;
 181
 182         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 183         if (!tbl.data)
 184                 return -ENOMEM;
 185         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 186         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 187         kfree(tbl.data);
 188         return ret;
 189 }
 190
 191 static int proc_allowed_congestion_control(ctl_table *ctl,
 192                                            int write,
 193                                            void __user *buffer, size_t *lenp,
 194                                            loff_t *ppos)
 195 {
 196         ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 197         int ret;
 198
 199         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 200         if (!tbl.data)
 201                 return -ENOMEM;
 202
 203         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 204         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 205         if (write && ret == 0)
 206                 ret = tcp_set_allowed_congestion_control(tbl.data);
 207         kfree(tbl.data);
 208         return ret;
 209 }
 210
 211 static int ipv4_tcp_mem(ctl_table *ctl, int write,
 212                            void __user *buffer, size_t *lenp,
 213                            loff_t *ppos)
 214 {
 215         int ret;
 216         unsigned long vec[3];
 217         struct net *net = current->nsproxy->net_ns;
 218 #ifdef CONFIG_MEMCG_KMEM
 219         struct mem_cgroup *memcg;
 220 #endif
 221
 222         ctl_table tmp = {
 223                 .data = &vec,
 224                 .maxlen = sizeof(vec),
 225                 .mode = ctl->mode,
 226         };
 227
 228         if (!write) {
 229                 ctl->data = &net->ipv4.sysctl_tcp_mem;
 230                 return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 231         }
 232
 233         ret = proc_doulongvec_minmax(&tmp, write, buffer, lenp, ppos);
 234         if (ret)
 235                 return ret;
 236
 237 #ifdef CONFIG_MEMCG_KMEM
 238         rcu_read_lock();
 239         memcg = mem_cgroup_from_task(current);
 240
 241         tcp_prot_mem(memcg, vec[0], 0);
 242         tcp_prot_mem(memcg, vec[1], 1);
 243         tcp_prot_mem(memcg, vec[2], 2);
 244         rcu_read_unlock();
 245 #endif
 246
 247         net->ipv4.sysctl_tcp_mem[0] = vec[0];
 248         net->ipv4.sysctl_tcp_mem[1] = vec[1];
 249         net->ipv4.sysctl_tcp_mem[2] = vec[2];
 250
 251         return 0;
 252 }
 253
 254 static int proc_tcp_fastopen_key(ctl_table *ctl, int write, void __user *buffer,
 255                                  size_t *lenp, loff_t *ppos)
 256 {
 257         ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 258         struct tcp_fastopen_context *ctxt;
 259         int ret;
 260         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 261
 262         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 263         if (!tbl.data)
 264                 return -ENOMEM;
 265
 266         rcu_read_lock();
 267         ctxt = rcu_dereference(tcp_fastopen_ctx);
 268         if (ctxt)
 269                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 270         else
 271                 memset(user_key, 0, sizeof(user_key));
 272         rcu_read_unlock();
 273
 274         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 275                 user_key[0], user_key[1], user_key[2], user_key[3]);
 276         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 277
 278         if (write && ret == 0) {
 279                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 280                            user_key + 2, user_key + 3) != 4) {
 281                         ret = -EINVAL;
 282                         goto bad_key;
 283                 }
 284                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 285         }
 286
 287 bad_key:
 288         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 289                user_key[0], user_key[1], user_key[2], user_key[3],
 290                (char *)tbl.data, ret);
 291         kfree(tbl.data);
 292         return ret;
 293 }
 294
 295 static struct ctl_table ipv4_table[] = {
 296         {
 297                 .procname       = "tcp_timestamps",
 298                 .data           = &sysctl_tcp_timestamps,
 299                 .maxlen         = sizeof(int),
 300                 .mode           = 0644,
 301                 .proc_handler   = proc_dointvec
 302         },
 303         {
 304                 .procname       = "tcp_window_scaling",
 305                 .data           = &sysctl_tcp_window_scaling,
 306                 .maxlen         = sizeof(int),
 307                 .mode           = 0644,
 308                 .proc_handler   = proc_dointvec
 309         },
 310         {
 311                 .procname       = "tcp_sack",
 312                 .data           = &sysctl_tcp_sack,
 313                 .maxlen         = sizeof(int),
 314                 .mode           = 0644,
 315                 .proc_handler   = proc_dointvec
 316         },
 317         {
 318                 .procname       = "tcp_retrans_collapse",
 319                 .data           = &sysctl_tcp_retrans_collapse,
 320                 .maxlen         = sizeof(int),
 321                 .mode           = 0644,
 322                 .proc_handler   = proc_dointvec
 323         },
 324         {
 325                 .procname       = "ip_default_ttl",
 326                 .data           = &sysctl_ip_default_ttl,
 327                 .maxlen         = sizeof(int),
 328                 .mode           = 0644,
 329                 .proc_handler   = proc_dointvec_minmax,
 330                 .extra1         = &ip_ttl_min,
 331                 .extra2         = &ip_ttl_max,
 332         },
 333         {
 334                 .procname       = "ip_no_pmtu_disc",
 335                 .data           = &ipv4_config.no_pmtu_disc,
 336                 .maxlen         = sizeof(int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec
 339         },
 340         {
 341                 .procname       = "ip_nonlocal_bind",
 342                 .data           = &sysctl_ip_nonlocal_bind,
 343                 .maxlen         = sizeof(int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec
 346         },
 347         {
 348                 .procname       = "tcp_syn_retries",
 349                 .data           = &sysctl_tcp_syn_retries,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec_minmax,
 353                 .extra1         = &tcp_syn_retries_min,
 354                 .extra2         = &tcp_syn_retries_max
 355         },
 356         {
 357                 .procname       = "tcp_synack_retries",
 358                 .data           = &sysctl_tcp_synack_retries,
 359                 .maxlen         = sizeof(int),
 360                 .mode           = 0644,
 361                 .proc_handler   = proc_dointvec
 362         },
 363         {
 364                 .procname       = "tcp_max_orphans",
 365                 .data           = &sysctl_tcp_max_orphans,
 366                 .maxlen         = sizeof(int),
 367                 .mode           = 0644,
 368                 .proc_handler   = proc_dointvec
 369         },
 370         {
 371                 .procname       = "tcp_max_tw_buckets",
 372                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 373                 .maxlen         = sizeof(int),
 374                 .mode           = 0644,
 375                 .proc_handler   = proc_dointvec
 376         },
 377         {
 378                 .procname       = "ip_early_demux",
 379                 .data           = &sysctl_ip_early_demux,
 380                 .maxlen         = sizeof(int),
 381                 .mode           = 0644,
 382                 .proc_handler   = proc_dointvec
 383         },
 384         {
 385                 .procname       = "ip_dynaddr",
 386                 .data           = &sysctl_ip_dynaddr,
 387                 .maxlen         = sizeof(int),
 388                 .mode           = 0644,
 389                 .proc_handler   = proc_dointvec
 390         },
 391         {
 392                 .procname       = "tcp_keepalive_time",
 393                 .data           = &sysctl_tcp_keepalive_time,
 394                 .maxlen         = sizeof(int),
 395                 .mode           = 0644,
 396                 .proc_handler   = proc_dointvec_jiffies,
 397         },
 398         {
 399                 .procname       = "tcp_keepalive_probes",
 400                 .data           = &sysctl_tcp_keepalive_probes,
 401                 .maxlen         = sizeof(int),
 402                 .mode           = 0644,
 403                 .proc_handler   = proc_dointvec
 404         },
 405         {
 406                 .procname       = "tcp_keepalive_intvl",
 407                 .data           = &sysctl_tcp_keepalive_intvl,
 408                 .maxlen         = sizeof(int),
 409                 .mode           = 0644,
 410                 .proc_handler   = proc_dointvec_jiffies,
 411         },
 412         {
 413                 .procname       = "tcp_retries1",
 414                 .data           = &sysctl_tcp_retries1,
 415                 .maxlen         = sizeof(int),
 416                 .mode           = 0644,
 417                 .proc_handler   = proc_dointvec_minmax,
 418                 .extra2         = &tcp_retr1_max
 419         },
 420         {
 421                 .procname       = "tcp_retries2",
 422                 .data           = &sysctl_tcp_retries2,
 423                 .maxlen         = sizeof(int),
 424                 .mode           = 0644,
 425                 .proc_handler   = proc_dointvec
 426         },
 427         {
 428                 .procname       = "tcp_fin_timeout",
 429                 .data           = &sysctl_tcp_fin_timeout,
 430                 .maxlen         = sizeof(int),
 431                 .mode           = 0644,
 432                 .proc_handler   = proc_dointvec_jiffies,
 433         },
 434 #ifdef CONFIG_SYN_COOKIES
 435         {
 436                 .procname       = "tcp_syncookies",
 437                 .data           = &sysctl_tcp_syncookies,
 438                 .maxlen         = sizeof(int),
 439                 .mode           = 0644,
 440                 .proc_handler   = proc_dointvec
 441         },
 442 #endif
 443         {
 444                 .procname       = "tcp_fastopen",
 445                 .data           = &sysctl_tcp_fastopen,
 446                 .maxlen         = sizeof(int),
 447                 .mode           = 0644,
 448                 .proc_handler   = proc_dointvec,
 449         },
 450         {
 451                 .procname       = "tcp_fastopen_key",
 452                 .mode           = 0600,
 453                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 454                 .proc_handler   = proc_tcp_fastopen_key,
 455         },
 456         {
 457                 .procname       = "tcp_tw_recycle",
 458                 .data           = &tcp_death_row.sysctl_tw_recycle,
 459                 .maxlen         = sizeof(int),
 460                 .mode           = 0644,
 461                 .proc_handler   = proc_dointvec
 462         },
 463         {
 464                 .procname       = "tcp_abort_on_overflow",
 465                 .data           = &sysctl_tcp_abort_on_overflow,
 466                 .maxlen         = sizeof(int),
 467                 .mode           = 0644,
 468                 .proc_handler   = proc_dointvec
 469         },
 470         {
 471                 .procname       = "tcp_stdurg",
 472                 .data           = &sysctl_tcp_stdurg,
 473                 .maxlen         = sizeof(int),
 474                 .mode           = 0644,
 475                 .proc_handler   = proc_dointvec
 476         },
 477         {
 478                 .procname       = "tcp_rfc1337",
 479                 .data           = &sysctl_tcp_rfc1337,
 480                 .maxlen         = sizeof(int),
 481                 .mode           = 0644,
 482                 .proc_handler   = proc_dointvec
 483         },
 484         {
 485                 .procname       = "tcp_max_syn_backlog",
 486                 .data           = &sysctl_max_syn_backlog,
 487                 .maxlen         = sizeof(int),
 488                 .mode           = 0644,
 489                 .proc_handler   = proc_dointvec
 490         },
 491         {
 492                 .procname       = "ip_local_port_range",
 493                 .data           = &sysctl_local_ports.range,
 494                 .maxlen         = sizeof(sysctl_local_ports.range),
 495                 .mode           = 0644,
 496                 .proc_handler   = ipv4_local_port_range,
 497         },
 498         {
 499                 .procname       = "ip_local_reserved_ports",
 500                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 501                 .maxlen         = 65536,
 502                 .mode           = 0644,
 503                 .proc_handler   = proc_do_large_bitmap,
 504         },
 505         {
 506                 .procname       = "igmp_max_memberships",
 507                 .data           = &sysctl_igmp_max_memberships,
 508                 .maxlen         = sizeof(int),
 509                 .mode           = 0644,
 510                 .proc_handler   = proc_dointvec
 511         },
 512         {
 513                 .procname       = "igmp_max_msf",
 514                 .data           = &sysctl_igmp_max_msf,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec
 518         },
 519         {
 520                 .procname       = "inet_peer_threshold",
 521                 .data           = &inet_peer_threshold,
 522                 .maxlen         = sizeof(int),
 523                 .mode           = 0644,
 524                 .proc_handler   = proc_dointvec
 525         },
 526         {
 527                 .procname       = "inet_peer_minttl",
 528                 .data           = &inet_peer_minttl,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = proc_dointvec_jiffies,
 532         },
 533         {
 534                 .procname       = "inet_peer_maxttl",
 535                 .data           = &inet_peer_maxttl,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = proc_dointvec_jiffies,
 539         },
 540         {
 541                 .procname       = "tcp_orphan_retries",
 542                 .data           = &sysctl_tcp_orphan_retries,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec
 546         },
 547         {
 548                 .procname       = "tcp_fack",
 549                 .data           = &sysctl_tcp_fack,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec
 553         },
 554         {
 555                 .procname       = "tcp_reordering",
 556                 .data           = &sysctl_tcp_reordering,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_dointvec
 560         },
 561         {
 562                 .procname       = "tcp_dsack",
 563                 .data           = &sysctl_tcp_dsack,
 564                 .maxlen         = sizeof(int),
 565                 .mode           = 0644,
 566                 .proc_handler   = proc_dointvec
 567         },
 568         {
 569                 .procname       = "tcp_wmem",
 570                 .data           = &sysctl_tcp_wmem,
 571                 .maxlen         = sizeof(sysctl_tcp_wmem),
 572                 .mode           = 0644,
 573                 .proc_handler   = proc_dointvec_minmax,
 574                 .extra1         = &one,
 575         },
 576         {
 577                 .procname       = "tcp_rmem",
 578                 .data           = &sysctl_tcp_rmem,
 579                 .maxlen         = sizeof(sysctl_tcp_rmem),
 580                 .mode           = 0644,
 581                 .proc_handler   = proc_dointvec_minmax,
 582                 .extra1         = &one,
 583         },
 584         {
 585                 .procname       = "tcp_app_win",
 586                 .data           = &sysctl_tcp_app_win,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec
 590         },
 591         {
 592                 .procname       = "tcp_adv_win_scale",
 593                 .data           = &sysctl_tcp_adv_win_scale,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec_minmax,
 597                 .extra1         = &tcp_adv_win_scale_min,
 598                 .extra2         = &tcp_adv_win_scale_max,
 599         },
 600         {
 601                 .procname       = "tcp_tw_reuse",
 602                 .data           = &sysctl_tcp_tw_reuse,
 603                 .maxlen         = sizeof(int),
 604                 .mode           = 0644,
 605                 .proc_handler   = proc_dointvec
 606         },
 607         {
 608                 .procname       = "tcp_frto",
 609                 .data           = &sysctl_tcp_frto,
 610                 .maxlen         = sizeof(int),
 611                 .mode           = 0644,
 612                 .proc_handler   = proc_dointvec
 613         },
 614         {
 615                 .procname       = "tcp_low_latency",
 616                 .data           = &sysctl_tcp_low_latency,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec
 620         },
 621         {
 622                 .procname       = "tcp_no_metrics_save",
 623                 .data           = &sysctl_tcp_nometrics_save,
 624                 .maxlen         = sizeof(int),
 625                 .mode           = 0644,
 626                 .proc_handler   = proc_dointvec,
 627         },
 628         {
 629                 .procname       = "tcp_moderate_rcvbuf",
 630                 .data           = &sysctl_tcp_moderate_rcvbuf,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec,
 634         },
 635         {
 636                 .procname       = "tcp_tso_win_divisor",
 637                 .data           = &sysctl_tcp_tso_win_divisor,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec,
 641         },
 642         {
 643                 .procname       = "tcp_congestion_control",
 644                 .mode           = 0644,
 645                 .maxlen         = TCP_CA_NAME_MAX,
 646                 .proc_handler   = proc_tcp_congestion_control,
 647         },
 648         {
 649                 .procname       = "tcp_mtu_probing",
 650                 .data           = &sysctl_tcp_mtu_probing,
 651                 .maxlen         = sizeof(int),
 652                 .mode           = 0644,
 653                 .proc_handler   = proc_dointvec,
 654         },
 655         {
 656                 .procname       = "tcp_base_mss",
 657                 .data           = &sysctl_tcp_base_mss,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec,
 661         },
 662         {
 663                 .procname       = "tcp_workaround_signed_windows",
 664                 .data           = &sysctl_tcp_workaround_signed_windows,
 665                 .maxlen         = sizeof(int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec
 668         },
 669         {
 670                 .procname       = "tcp_limit_output_bytes",
 671                 .data           = &sysctl_tcp_limit_output_bytes,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec
 675         },
 676         {
 677                 .procname       = "tcp_challenge_ack_limit",
 678                 .data           = &sysctl_tcp_challenge_ack_limit,
 679                 .maxlen         = sizeof(int),
 680                 .mode           = 0644,
 681                 .proc_handler   = proc_dointvec
 682         },
 683 #ifdef CONFIG_NET_DMA
 684         {
 685                 .procname       = "tcp_dma_copybreak",
 686                 .data           = &sysctl_tcp_dma_copybreak,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec
 690         },
 691 #endif
 692         {
 693                 .procname       = "tcp_slow_start_after_idle",
 694                 .data           = &sysctl_tcp_slow_start_after_idle,
 695                 .maxlen         = sizeof(int),
 696                 .mode           = 0644,
 697                 .proc_handler   = proc_dointvec
 698         },
 699 #ifdef CONFIG_NETLABEL
 700         {
 701                 .procname       = "cipso_cache_enable",
 702                 .data           = &cipso_v4_cache_enabled,
 703                 .maxlen         = sizeof(int),
 704                 .mode           = 0644,
 705                 .proc_handler   = proc_dointvec,
 706         },
 707         {
 708                 .procname       = "cipso_cache_bucket_size",
 709                 .data           = &cipso_v4_cache_bucketsize,
 710                 .maxlen         = sizeof(int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec,
 713         },
 714         {
 715                 .procname       = "cipso_rbm_optfmt",
 716                 .data           = &cipso_v4_rbm_optfmt,
 717                 .maxlen         = sizeof(int),
 718                 .mode           = 0644,
 719                 .proc_handler   = proc_dointvec,
 720         },
 721         {
 722                 .procname       = "cipso_rbm_strictvalid",
 723                 .data           = &cipso_v4_rbm_strictvalid,
 724                 .maxlen         = sizeof(int),
 725                 .mode           = 0644,
 726                 .proc_handler   = proc_dointvec,
 727         },
 728 #endif /* CONFIG_NETLABEL */
 729         {
 730                 .procname       = "tcp_available_congestion_control",
 731                 .maxlen         = TCP_CA_BUF_MAX,
 732                 .mode           = 0444,
 733                 .proc_handler   = proc_tcp_available_congestion_control,
 734         },
 735         {
 736                 .procname       = "tcp_allowed_congestion_control",
 737                 .maxlen         = TCP_CA_BUF_MAX,
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_allowed_congestion_control,
 740         },
 741         {
 742                 .procname       = "tcp_max_ssthresh",
 743                 .data           = &sysctl_tcp_max_ssthresh,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dointvec,
 747         },
 748         {
 749                 .procname       = "tcp_thin_linear_timeouts",
 750                 .data           = &sysctl_tcp_thin_linear_timeouts,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec
 754         },
 755         {
 756                 .procname       = "tcp_thin_dupack",
 757                 .data           = &sysctl_tcp_thin_dupack,
 758                 .maxlen         = sizeof(int),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_dointvec
 761         },
 762         {
 763                 .procname       = "tcp_early_retrans",
 764                 .data           = &sysctl_tcp_early_retrans,
 765                 .maxlen         = sizeof(int),
 766                 .mode           = 0644,
 767                 .proc_handler   = proc_dointvec_minmax,
 768                 .extra1         = &zero,
 769                 .extra2         = &four,
 770         },
 771         {
 772                 .procname       = "tcp_min_tso_segs",
 773                 .data           = &sysctl_tcp_min_tso_segs,
 774                 .maxlen         = sizeof(int),
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_dointvec_minmax,
 777                 .extra1         = &zero,
 778                 .extra2         = &gso_max_segs,
 779         },
 780         {
 781                 .procname       = "tcp_default_init_rwnd",
 782                 .data           = &sysctl_tcp_default_init_rwnd,
 783                 .maxlen         = sizeof(int),
 784                 .mode           = 0644,
 785                 .proc_handler   = proc_tcp_default_init_rwnd
 786         },
 787         {
 788                 .procname       = "udp_mem",
 789                 .data           = &sysctl_udp_mem,
 790                 .maxlen         = sizeof(sysctl_udp_mem),
 791                 .mode           = 0644,
 792                 .proc_handler   = proc_doulongvec_minmax,
 793         },
 794         {
 795                 .procname       = "udp_rmem_min",
 796                 .data           = &sysctl_udp_rmem_min,
 797                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 798                 .mode           = 0644,
 799                 .proc_handler   = proc_dointvec_minmax,
 800                 .extra1         = &one
 801         },
 802         {
 803                 .procname       = "udp_wmem_min",
 804                 .data           = &sysctl_udp_wmem_min,
 805                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 806                 .mode           = 0644,
 807                 .proc_handler   = proc_dointvec_minmax,
 808                 .extra1         = &one
 809         },
 810         { }
 811 };
 812
 813 static struct ctl_table ipv4_net_table[] = {
 814         {
 815                 .procname       = "icmp_echo_ignore_all",
 816                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 817                 .maxlen         = sizeof(int),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_dointvec
 820         },
 821         {
 822                 .procname       = "icmp_echo_ignore_broadcasts",
 823                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 824                 .maxlen         = sizeof(int),
 825                 .mode           = 0644,
 826                 .proc_handler   = proc_dointvec
 827         },
 828         {
 829                 .procname       = "icmp_ignore_bogus_error_responses",
 830                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 831                 .maxlen         = sizeof(int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec
 834         },
 835         {
 836                 .procname       = "icmp_errors_use_inbound_ifaddr",
 837                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 838                 .maxlen         = sizeof(int),
 839                 .mode           = 0644,
 840                 .proc_handler   = proc_dointvec
 841         },
 842         {
 843                 .procname       = "icmp_ratelimit",
 844                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 845                 .maxlen         = sizeof(int),
 846                 .mode           = 0644,
 847                 .proc_handler   = proc_dointvec_ms_jiffies,
 848         },
 849         {
 850                 .procname       = "icmp_ratemask",
 851                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 852                 .maxlen         = sizeof(int),
 853                 .mode           = 0644,
 854                 .proc_handler   = proc_dointvec
 855         },
 856         {
 857                 .procname       = "ping_group_range",
 858                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 859                 .maxlen         = sizeof(gid_t)*2,
 860                 .mode           = 0644,
 861                 .proc_handler   = ipv4_ping_group_range,
 862         },
 863         {
 864                 .procname       = "tcp_ecn",
 865                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 866                 .maxlen         = sizeof(int),
 867                 .mode           = 0644,
 868                 .proc_handler   = proc_dointvec
 869         },
 870         {
 871                 .procname       = "tcp_mem",
 872                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 873                 .mode           = 0644,
 874                 .proc_handler   = ipv4_tcp_mem,
 875         },
 876         {
 877                 .procname       = "fwmark_reflect",
 878                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 879                 .maxlen         = sizeof(int),
 880                 .mode           = 0644,
 881                 .proc_handler   = proc_dointvec,
 882         },
 883         {
 884                 .procname       = "tcp_fwmark_accept",
 885                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 886                 .maxlen         = sizeof(int),
 887                 .mode           = 0644,
 888                 .proc_handler   = proc_dointvec,
 889         },
 890         { }
 891 };
 892
 893 static __net_init int ipv4_sysctl_init_net(struct net *net)
 894 {
 895         struct ctl_table *table;
 896
 897         table = ipv4_net_table;
 898         if (!net_eq(net, &init_net)) {
 899                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 900                 if (table == NULL)
 901                         goto err_alloc;
 902
 903                 table[0].data =
 904                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 905                 table[1].data =
 906                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 907                 table[2].data =
 908                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 909                 table[3].data =
 910                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 911                 table[4].data =
 912                         &net->ipv4.sysctl_icmp_ratelimit;
 913                 table[5].data =
 914                         &net->ipv4.sysctl_icmp_ratemask;
 915                 table[6].data =
 916                         &net->ipv4.sysctl_ping_group_range;
 917                 table[7].data =
 918                         &net->ipv4.sysctl_tcp_ecn;
 919
 920                 /* Don't export sysctls to unprivileged users */
 921                 if (net->user_ns != &init_user_ns)
 922                         table[0].procname = NULL;
 923         }
 924
 925         /*
 926          * Sane defaults - nobody may create ping sockets.
 927          * Boot scripts should set this to distro-specific group.
 928          */
 929         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 930         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 931
 932         tcp_init_mem(net);
 933
 934         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 935         if (net->ipv4.ipv4_hdr == NULL)
 936                 goto err_reg;
 937
 938         return 0;
 939
 940 err_reg:
 941         if (!net_eq(net, &init_net))
 942                 kfree(table);
 943 err_alloc:
 944         return -ENOMEM;
 945 }
 946
 947 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 948 {
 949         struct ctl_table *table;
 950
 951         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 952         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 953         kfree(table);
 954 }
 955
 956 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 957         .init = ipv4_sysctl_init_net,
 958         .exit = ipv4_sysctl_exit_net,
 959 };
 960
 961 static __init int sysctl_ipv4_init(void)
 962 {
 963         struct ctl_table_header *hdr;
 964         struct ctl_table *i;
 965
 966         for (i = ipv4_table; i->procname; i++) {
 967                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 968                         i->data = sysctl_local_reserved_ports;
 969                         break;
 970                 }
 971         }
 972         if (!i->procname)
 973                 return -EINVAL;
 974
 975         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 976         if (hdr == NULL)
 977                 return -ENOMEM;
 978
 979         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 980                 unregister_net_sysctl_table(hdr);
 981                 return -ENOMEM;
 982         }
 983
 984         return 0;
 985 }
 986
 987 __initcall(sysctl_ipv4_init);