1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem. 4 * 5 * Begun April 1, 1996, Mike Shaver. 6 * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS] 7 */ 8 9 #include <linux/mm.h> 10 #include <linux/module.h> 11 #include <linux/sysctl.h> 12 #include <linux/igmp.h> 13 #include <linux/inetdevice.h> 14 #include <linux/seqlock.h> 15 #include <linux/init.h> 16 #include <linux/slab.h> 17 #include <linux/nsproxy.h> 18 #include <linux/swap.h> 19 #include <net/snmp.h> 20 #include <net/icmp.h> 21 #include <net/ip.h> 22 #include <net/route.h> 23 #include <net/tcp.h> 24 #include <net/udp.h> 25 #include <net/cipso_ipv4.h> 26 #include <net/inet_frag.h> 27 #include <net/ping.h> 28 #include <net/protocol.h> 29 #include <net/netevent.h> 30 31 static int zero; 32 static int one = 1; 33 static int two = 2; 34 static int four = 4; 35 static int thousand = 1000; 36 static int gso_max_segs = GSO_MAX_SEGS; 37 static int tcp_retr1_max = 255; 38 static int ip_local_port_range_min[] = { 1, 1 }; 39 static int ip_local_port_range_max[] = { 65535, 65535 }; 40 static int tcp_adv_win_scale_min = -31; 41 static int tcp_adv_win_scale_max = 31; 42 static int ip_privileged_port_min; 43 static int ip_privileged_port_max = 65535; 44 static int ip_ttl_min = 1; 45 static int ip_ttl_max = 255; 46 static int tcp_syn_retries_min = 1; 47 static int tcp_syn_retries_max = MAX_TCP_SYNCNT; 48 static int ip_ping_group_range_min[] = { 0, 0 }; 49 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX }; 50 static int comp_sack_nr_max = 255; 51 static u32 u32_max_div_HZ = UINT_MAX / HZ; 52 static int one_day_secs = 24 * 3600; 53 54 /* obsolete */ 55 static int sysctl_tcp_low_latency __read_mostly; 56 57 /* Update system visible IP port range */ 58 static void set_local_port_range(struct net *net, int range[2]) 59 { 60 bool same_parity = !((range[0] ^ range[1]) & 1); 61 62 write_seqlock_bh(&net->ipv4.ip_local_ports.lock); 63 if (same_parity && !net->ipv4.ip_local_ports.warned) { 64 net->ipv4.ip_local_ports.warned = true; 65 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n"); 66 } 67 net->ipv4.ip_local_ports.range[0] = range[0]; 68 net->ipv4.ip_local_ports.range[1] = range[1]; 69 write_sequnlock_bh(&net->ipv4.ip_local_ports.lock); 70 } 71 72 /* Validate changes from /proc interface. */ 73 static int ipv4_local_port_range(struct ctl_table *table, int write, 74 void __user *buffer, 75 size_t *lenp, loff_t *ppos) 76 { 77 struct net *net = 78 container_of(table->data, struct net, ipv4.ip_local_ports.range); 79 int ret; 80 int range[2]; 81 struct ctl_table tmp = { 82 .data = &range, 83 .maxlen = sizeof(range), 84 .mode = table->mode, 85 .extra1 = &ip_local_port_range_min, 86 .extra2 = &ip_local_port_range_max, 87 }; 88 89 inet_get_local_port_range(net, &range[0], &range[1]); 90 91 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); 92 93 if (write && ret == 0) { 94 /* Ensure that the upper limit is not smaller than the lower, 95 * and that the lower does not encroach upon the privileged 96 * port limit. 97 */ 98 if ((range[1] < range[0]) || 99 (range[0] < net->ipv4.sysctl_ip_prot_sock)) 100 ret = -EINVAL; 101 else 102 set_local_port_range(net, range); 103 } 104 105 return ret; 106 } 107 108 /* Validate changes from /proc interface. */ 109 static int ipv4_privileged_ports(struct ctl_table *table, int write, 110 void __user *buffer, size_t *lenp, loff_t *ppos) 111 { 112 struct net *net = container_of(table->data, struct net, 113 ipv4.sysctl_ip_prot_sock); 114 int ret; 115 int pports; 116 int range[2]; 117 struct ctl_table tmp = { 118 .data = &pports, 119 .maxlen = sizeof(pports), 120 .mode = table->mode, 121 .extra1 = &ip_privileged_port_min, 122 .extra2 = &ip_privileged_port_max, 123 }; 124 125 pports = net->ipv4.sysctl_ip_prot_sock; 126 127 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); 128 129 if (write && ret == 0) { 130 inet_get_local_port_range(net, &range[0], &range[1]); 131 /* Ensure that the local port range doesn't overlap with the 132 * privileged port range. 133 */ 134 if (range[0] < pports) 135 ret = -EINVAL; 136 else 137 net->ipv4.sysctl_ip_prot_sock = pports; 138 } 139 140 return ret; 141 } 142 143 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high) 144 { 145 kgid_t *data = table->data; 146 struct net *net = 147 container_of(table->data, struct net, ipv4.ping_group_range.range); 148 unsigned int seq; 149 do { 150 seq = read_seqbegin(&net->ipv4.ping_group_range.lock); 151 152 *low = data[0]; 153 *high = data[1]; 154 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq)); 155 } 156 157 /* Update system visible IP port range */ 158 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high) 159 { 160 kgid_t *data = table->data; 161 struct net *net = 162 container_of(table->data, struct net, ipv4.ping_group_range.range); 163 write_seqlock(&net->ipv4.ping_group_range.lock); 164 data[0] = low; 165 data[1] = high; 166 write_sequnlock(&net->ipv4.ping_group_range.lock); 167 } 168 169 /* Validate changes from /proc interface. */ 170 static int ipv4_ping_group_range(struct ctl_table *table, int write, 171 void __user *buffer, 172 size_t *lenp, loff_t *ppos) 173 { 174 struct user_namespace *user_ns = current_user_ns(); 175 int ret; 176 gid_t urange[2]; 177 kgid_t low, high; 178 struct ctl_table tmp = { 179 .data = &urange, 180 .maxlen = sizeof(urange), 181 .mode = table->mode, 182 .extra1 = &ip_ping_group_range_min, 183 .extra2 = &ip_ping_group_range_max, 184 }; 185 186 inet_get_ping_group_range_table(table, &low, &high); 187 urange[0] = from_kgid_munged(user_ns, low); 188 urange[1] = from_kgid_munged(user_ns, high); 189 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); 190 191 if (write && ret == 0) { 192 low = make_kgid(user_ns, urange[0]); 193 high = make_kgid(user_ns, urange[1]); 194 if (!gid_valid(low) || !gid_valid(high)) 195 return -EINVAL; 196 if (urange[1] < urange[0] || gid_lt(high, low)) { 197 low = make_kgid(&init_user_ns, 1); 198 high = make_kgid(&init_user_ns, 0); 199 } 200 set_ping_group_range(table, low, high); 201 } 202 203 return ret; 204 } 205 206 static int ipv4_fwd_update_priority(struct ctl_table *table, int write, 207 void __user *buffer, 208 size_t *lenp, loff_t *ppos) 209 { 210 struct net *net; 211 int ret; 212 213 net = container_of(table->data, struct net, 214 ipv4.sysctl_ip_fwd_update_priority); 215 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); 216 if (write && ret == 0) 217 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE, 218 net); 219 220 return ret; 221 } 222 223 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write, 224 void __user *buffer, size_t *lenp, loff_t *ppos) 225 { 226 struct net *net = container_of(ctl->data, struct net, 227 ipv4.tcp_congestion_control); 228 char val[TCP_CA_NAME_MAX]; 229 struct ctl_table tbl = { 230 .data = val, 231 .maxlen = TCP_CA_NAME_MAX, 232 }; 233 int ret; 234 235 tcp_get_default_congestion_control(net, val); 236 237 ret = proc_dostring(&tbl, write, buffer, lenp, ppos); 238 if (write && ret == 0) 239 ret = tcp_set_default_congestion_control(net, val); 240 return ret; 241 } 242 243 static int proc_tcp_available_congestion_control(struct ctl_table *ctl, 244 int write, 245 void __user *buffer, size_t *lenp, 246 loff_t *ppos) 247 { 248 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, }; 249 int ret; 250 251 tbl.data = kmalloc(tbl.maxlen, GFP_USER); 252 if (!tbl.data) 253 return -ENOMEM; 254 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX); 255 ret = proc_dostring(&tbl, write, buffer, lenp, ppos); 256 kfree(tbl.data); 257 return ret; 258 } 259 260 static int proc_allowed_congestion_control(struct ctl_table *ctl, 261 int write, 262 void __user *buffer, size_t *lenp, 263 loff_t *ppos) 264 { 265 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX }; 266 int ret; 267 268 tbl.data = kmalloc(tbl.maxlen, GFP_USER); 269 if (!tbl.data) 270 return -ENOMEM; 271 272 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen); 273 ret = proc_dostring(&tbl, write, buffer, lenp, ppos); 274 if (write && ret == 0) 275 ret = tcp_set_allowed_congestion_control(tbl.data); 276 kfree(tbl.data); 277 return ret; 278 } 279 280 static int proc_tcp_fastopen_key(struct ctl_table *table, int write, 281 void __user *buffer, size_t *lenp, 282 loff_t *ppos) 283 { 284 struct net *net = container_of(table->data, struct net, 285 ipv4.sysctl_tcp_fastopen); 286 struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) }; 287 struct tcp_fastopen_context *ctxt; 288 u32 user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */ 289 __le32 key[4]; 290 int ret, i; 291 292 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL); 293 if (!tbl.data) 294 return -ENOMEM; 295 296 rcu_read_lock(); 297 ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx); 298 if (ctxt) 299 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH); 300 else 301 memset(key, 0, sizeof(key)); 302 rcu_read_unlock(); 303 304 for (i = 0; i < ARRAY_SIZE(key); i++) 305 user_key[i] = le32_to_cpu(key[i]); 306 307 snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x", 308 user_key[0], user_key[1], user_key[2], user_key[3]); 309 ret = proc_dostring(&tbl, write, buffer, lenp, ppos); 310 311 if (write && ret == 0) { 312 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1, 313 user_key + 2, user_key + 3) != 4) { 314 ret = -EINVAL; 315 goto bad_key; 316 } 317 318 for (i = 0; i < ARRAY_SIZE(user_key); i++) 319 key[i] = cpu_to_le32(user_key[i]); 320 321 tcp_fastopen_reset_cipher(net, NULL, key, 322 TCP_FASTOPEN_KEY_LENGTH); 323 } 324 325 bad_key: 326 pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n", 327 user_key[0], user_key[1], user_key[2], user_key[3], 328 (char *)tbl.data, ret); 329 kfree(tbl.data); 330 return ret; 331 } 332 333 static void proc_configure_early_demux(int enabled, int protocol) 334 { 335 struct net_protocol *ipprot; 336 #if IS_ENABLED(CONFIG_IPV6) 337 struct inet6_protocol *ip6prot; 338 #endif 339 340 rcu_read_lock(); 341 342 ipprot = rcu_dereference(inet_protos[protocol]); 343 if (ipprot) 344 ipprot->early_demux = enabled ? ipprot->early_demux_handler : 345 NULL; 346 347 #if IS_ENABLED(CONFIG_IPV6) 348 ip6prot = rcu_dereference(inet6_protos[protocol]); 349 if (ip6prot) 350 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler : 351 NULL; 352 #endif 353 rcu_read_unlock(); 354 } 355 356 static int proc_tcp_early_demux(struct ctl_table *table, int write, 357 void __user *buffer, size_t *lenp, loff_t *ppos) 358 { 359 int ret = 0; 360 361 ret = proc_dointvec(table, write, buffer, lenp, ppos); 362 363 if (write && !ret) { 364 int enabled = init_net.ipv4.sysctl_tcp_early_demux; 365 366 proc_configure_early_demux(enabled, IPPROTO_TCP); 367 } 368 369 return ret; 370 } 371 372 static int proc_udp_early_demux(struct ctl_table *table, int write, 373 void __user *buffer, size_t *lenp, loff_t *ppos) 374 { 375 int ret = 0; 376 377 ret = proc_dointvec(table, write, buffer, lenp, ppos); 378 379 if (write && !ret) { 380 int enabled = init_net.ipv4.sysctl_udp_early_demux; 381 382 proc_configure_early_demux(enabled, IPPROTO_UDP); 383 } 384 385 return ret; 386 } 387 388 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table, 389 int write, 390 void __user *buffer, 391 size_t *lenp, loff_t *ppos) 392 { 393 struct net *net = container_of(table->data, struct net, 394 ipv4.sysctl_tcp_fastopen_blackhole_timeout); 395 int ret; 396 397 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); 398 if (write && ret == 0) 399 atomic_set(&net->ipv4.tfo_active_disable_times, 0); 400 401 return ret; 402 } 403 404 static int proc_tcp_available_ulp(struct ctl_table *ctl, 405 int write, 406 void __user *buffer, size_t *lenp, 407 loff_t *ppos) 408 { 409 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, }; 410 int ret; 411 412 tbl.data = kmalloc(tbl.maxlen, GFP_USER); 413 if (!tbl.data) 414 return -ENOMEM; 415 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX); 416 ret = proc_dostring(&tbl, write, buffer, lenp, ppos); 417 kfree(tbl.data); 418 419 return ret; 420 } 421 422 #ifdef CONFIG_IP_ROUTE_MULTIPATH 423 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write, 424 void __user *buffer, size_t *lenp, 425 loff_t *ppos) 426 { 427 struct net *net = container_of(table->data, struct net, 428 ipv4.sysctl_fib_multipath_hash_policy); 429 int ret; 430 431 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); 432 if (write && ret == 0) 433 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net); 434 435 return ret; 436 } 437 #endif 438 439 static struct ctl_table ipv4_table[] = { 440 { 441 .procname = "tcp_max_orphans", 442 .data = &sysctl_tcp_max_orphans, 443 .maxlen = sizeof(int), 444 .mode = 0644, 445 .proc_handler = proc_dointvec 446 }, 447 { 448 .procname = "inet_peer_threshold", 449 .data = &inet_peer_threshold, 450 .maxlen = sizeof(int), 451 .mode = 0644, 452 .proc_handler = proc_dointvec 453 }, 454 { 455 .procname = "inet_peer_minttl", 456 .data = &inet_peer_minttl, 457 .maxlen = sizeof(int), 458 .mode = 0644, 459 .proc_handler = proc_dointvec_jiffies, 460 }, 461 { 462 .procname = "inet_peer_maxttl", 463 .data = &inet_peer_maxttl, 464 .maxlen = sizeof(int), 465 .mode = 0644, 466 .proc_handler = proc_dointvec_jiffies, 467 }, 468 { 469 .procname = "tcp_mem", 470 .maxlen = sizeof(sysctl_tcp_mem), 471 .data = &sysctl_tcp_mem, 472 .mode = 0644, 473 .proc_handler = proc_doulongvec_minmax, 474 }, 475 { 476 .procname = "tcp_low_latency", 477 .data = &sysctl_tcp_low_latency, 478 .maxlen = sizeof(int), 479 .mode = 0644, 480 .proc_handler = proc_dointvec 481 }, 482 #ifdef CONFIG_NETLABEL 483 { 484 .procname = "cipso_cache_enable", 485 .data = &cipso_v4_cache_enabled, 486 .maxlen = sizeof(int), 487 .mode = 0644, 488 .proc_handler = proc_dointvec, 489 }, 490 { 491 .procname = "cipso_cache_bucket_size", 492 .data = &cipso_v4_cache_bucketsize, 493 .maxlen = sizeof(int), 494 .mode = 0644, 495 .proc_handler = proc_dointvec, 496 }, 497 { 498 .procname = "cipso_rbm_optfmt", 499 .data = &cipso_v4_rbm_optfmt, 500 .maxlen = sizeof(int), 501 .mode = 0644, 502 .proc_handler = proc_dointvec, 503 }, 504 { 505 .procname = "cipso_rbm_strictvalid", 506 .data = &cipso_v4_rbm_strictvalid, 507 .maxlen = sizeof(int), 508 .mode = 0644, 509 .proc_handler = proc_dointvec, 510 }, 511 #endif /* CONFIG_NETLABEL */ 512 { 513 .procname = "tcp_available_congestion_control", 514 .maxlen = TCP_CA_BUF_MAX, 515 .mode = 0444, 516 .proc_handler = proc_tcp_available_congestion_control, 517 }, 518 { 519 .procname = "tcp_allowed_congestion_control", 520 .maxlen = TCP_CA_BUF_MAX, 521 .mode = 0644, 522 .proc_handler = proc_allowed_congestion_control, 523 }, 524 { 525 .procname = "tcp_available_ulp", 526 .maxlen = TCP_ULP_BUF_MAX, 527 .mode = 0444, 528 .proc_handler = proc_tcp_available_ulp, 529 }, 530 { 531 .procname = "icmp_msgs_per_sec", 532 .data = &sysctl_icmp_msgs_per_sec, 533 .maxlen = sizeof(int), 534 .mode = 0644, 535 .proc_handler = proc_dointvec_minmax, 536 .extra1 = &zero, 537 }, 538 { 539 .procname = "icmp_msgs_burst", 540 .data = &sysctl_icmp_msgs_burst, 541 .maxlen = sizeof(int), 542 .mode = 0644, 543 .proc_handler = proc_dointvec_minmax, 544 .extra1 = &zero, 545 }, 546 { 547 .procname = "udp_mem", 548 .data = &sysctl_udp_mem, 549 .maxlen = sizeof(sysctl_udp_mem), 550 .mode = 0644, 551 .proc_handler = proc_doulongvec_minmax, 552 }, 553 { 554 .procname = "fib_sync_mem", 555 .data = &sysctl_fib_sync_mem, 556 .maxlen = sizeof(sysctl_fib_sync_mem), 557 .mode = 0644, 558 .proc_handler = proc_douintvec_minmax, 559 .extra1 = &sysctl_fib_sync_mem_min, 560 .extra2 = &sysctl_fib_sync_mem_max, 561 }, 562 { } 563 }; 564 565 static struct ctl_table ipv4_net_table[] = { 566 { 567 .procname = "icmp_echo_ignore_all", 568 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all, 569 .maxlen = sizeof(int), 570 .mode = 0644, 571 .proc_handler = proc_dointvec 572 }, 573 { 574 .procname = "icmp_echo_ignore_broadcasts", 575 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts, 576 .maxlen = sizeof(int), 577 .mode = 0644, 578 .proc_handler = proc_dointvec 579 }, 580 { 581 .procname = "icmp_ignore_bogus_error_responses", 582 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses, 583 .maxlen = sizeof(int), 584 .mode = 0644, 585 .proc_handler = proc_dointvec 586 }, 587 { 588 .procname = "icmp_errors_use_inbound_ifaddr", 589 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr, 590 .maxlen = sizeof(int), 591 .mode = 0644, 592 .proc_handler = proc_dointvec 593 }, 594 { 595 .procname = "icmp_ratelimit", 596 .data = &init_net.ipv4.sysctl_icmp_ratelimit, 597 .maxlen = sizeof(int), 598 .mode = 0644, 599 .proc_handler = proc_dointvec_ms_jiffies, 600 }, 601 { 602 .procname = "icmp_ratemask", 603 .data = &init_net.ipv4.sysctl_icmp_ratemask, 604 .maxlen = sizeof(int), 605 .mode = 0644, 606 .proc_handler = proc_dointvec 607 }, 608 { 609 .procname = "ping_group_range", 610 .data = &init_net.ipv4.ping_group_range.range, 611 .maxlen = sizeof(gid_t)*2, 612 .mode = 0644, 613 .proc_handler = ipv4_ping_group_range, 614 }, 615 #ifdef CONFIG_NET_L3_MASTER_DEV 616 { 617 .procname = "raw_l3mdev_accept", 618 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept, 619 .maxlen = sizeof(int), 620 .mode = 0644, 621 .proc_handler = proc_dointvec_minmax, 622 .extra1 = &zero, 623 .extra2 = &one, 624 }, 625 #endif 626 { 627 .procname = "tcp_ecn", 628 .data = &init_net.ipv4.sysctl_tcp_ecn, 629 .maxlen = sizeof(int), 630 .mode = 0644, 631 .proc_handler = proc_dointvec 632 }, 633 { 634 .procname = "tcp_ecn_fallback", 635 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback, 636 .maxlen = sizeof(int), 637 .mode = 0644, 638 .proc_handler = proc_dointvec 639 }, 640 { 641 .procname = "ip_dynaddr", 642 .data = &init_net.ipv4.sysctl_ip_dynaddr, 643 .maxlen = sizeof(int), 644 .mode = 0644, 645 .proc_handler = proc_dointvec 646 }, 647 { 648 .procname = "ip_early_demux", 649 .data = &init_net.ipv4.sysctl_ip_early_demux, 650 .maxlen = sizeof(int), 651 .mode = 0644, 652 .proc_handler = proc_dointvec 653 }, 654 { 655 .procname = "udp_early_demux", 656 .data = &init_net.ipv4.sysctl_udp_early_demux, 657 .maxlen = sizeof(int), 658 .mode = 0644, 659 .proc_handler = proc_udp_early_demux 660 }, 661 { 662 .procname = "tcp_early_demux", 663 .data = &init_net.ipv4.sysctl_tcp_early_demux, 664 .maxlen = sizeof(int), 665 .mode = 0644, 666 .proc_handler = proc_tcp_early_demux 667 }, 668 { 669 .procname = "ip_default_ttl", 670 .data = &init_net.ipv4.sysctl_ip_default_ttl, 671 .maxlen = sizeof(int), 672 .mode = 0644, 673 .proc_handler = proc_dointvec_minmax, 674 .extra1 = &ip_ttl_min, 675 .extra2 = &ip_ttl_max, 676 }, 677 { 678 .procname = "ip_local_port_range", 679 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range), 680 .data = &init_net.ipv4.ip_local_ports.range, 681 .mode = 0644, 682 .proc_handler = ipv4_local_port_range, 683 }, 684 { 685 .procname = "ip_local_reserved_ports", 686 .data = &init_net.ipv4.sysctl_local_reserved_ports, 687 .maxlen = 65536, 688 .mode = 0644, 689 .proc_handler = proc_do_large_bitmap, 690 }, 691 { 692 .procname = "ip_no_pmtu_disc", 693 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc, 694 .maxlen = sizeof(int), 695 .mode = 0644, 696 .proc_handler = proc_dointvec 697 }, 698 { 699 .procname = "ip_forward_use_pmtu", 700 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu, 701 .maxlen = sizeof(int), 702 .mode = 0644, 703 .proc_handler = proc_dointvec, 704 }, 705 { 706 .procname = "ip_forward_update_priority", 707 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority, 708 .maxlen = sizeof(int), 709 .mode = 0644, 710 .proc_handler = ipv4_fwd_update_priority, 711 .extra1 = &zero, 712 .extra2 = &one, 713 }, 714 { 715 .procname = "ip_nonlocal_bind", 716 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind, 717 .maxlen = sizeof(int), 718 .mode = 0644, 719 .proc_handler = proc_dointvec 720 }, 721 { 722 .procname = "fwmark_reflect", 723 .data = &init_net.ipv4.sysctl_fwmark_reflect, 724 .maxlen = sizeof(int), 725 .mode = 0644, 726 .proc_handler = proc_dointvec, 727 }, 728 { 729 .procname = "tcp_fwmark_accept", 730 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept, 731 .maxlen = sizeof(int), 732 .mode = 0644, 733 .proc_handler = proc_dointvec, 734 }, 735 #ifdef CONFIG_NET_L3_MASTER_DEV 736 { 737 .procname = "tcp_l3mdev_accept", 738 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept, 739 .maxlen = sizeof(int), 740 .mode = 0644, 741 .proc_handler = proc_dointvec_minmax, 742 .extra1 = &zero, 743 .extra2 = &one, 744 }, 745 #endif 746 { 747 .procname = "tcp_mtu_probing", 748 .data = &init_net.ipv4.sysctl_tcp_mtu_probing, 749 .maxlen = sizeof(int), 750 .mode = 0644, 751 .proc_handler = proc_dointvec, 752 }, 753 { 754 .procname = "tcp_base_mss", 755 .data = &init_net.ipv4.sysctl_tcp_base_mss, 756 .maxlen = sizeof(int), 757 .mode = 0644, 758 .proc_handler = proc_dointvec, 759 }, 760 { 761 .procname = "tcp_probe_threshold", 762 .data = &init_net.ipv4.sysctl_tcp_probe_threshold, 763 .maxlen = sizeof(int), 764 .mode = 0644, 765 .proc_handler = proc_dointvec, 766 }, 767 { 768 .procname = "tcp_probe_interval", 769 .data = &init_net.ipv4.sysctl_tcp_probe_interval, 770 .maxlen = sizeof(u32), 771 .mode = 0644, 772 .proc_handler = proc_douintvec_minmax, 773 .extra2 = &u32_max_div_HZ, 774 }, 775 { 776 .procname = "igmp_link_local_mcast_reports", 777 .data = &init_net.ipv4.sysctl_igmp_llm_reports, 778 .maxlen = sizeof(int), 779 .mode = 0644, 780 .proc_handler = proc_dointvec 781 }, 782 { 783 .procname = "igmp_max_memberships", 784 .data = &init_net.ipv4.sysctl_igmp_max_memberships, 785 .maxlen = sizeof(int), 786 .mode = 0644, 787 .proc_handler = proc_dointvec 788 }, 789 { 790 .procname = "igmp_max_msf", 791 .data = &init_net.ipv4.sysctl_igmp_max_msf, 792 .maxlen = sizeof(int), 793 .mode = 0644, 794 .proc_handler = proc_dointvec 795 }, 796 #ifdef CONFIG_IP_MULTICAST 797 { 798 .procname = "igmp_qrv", 799 .data = &init_net.ipv4.sysctl_igmp_qrv, 800 .maxlen = sizeof(int), 801 .mode = 0644, 802 .proc_handler = proc_dointvec_minmax, 803 .extra1 = &one 804 }, 805 #endif 806 { 807 .procname = "tcp_congestion_control", 808 .data = &init_net.ipv4.tcp_congestion_control, 809 .mode = 0644, 810 .maxlen = TCP_CA_NAME_MAX, 811 .proc_handler = proc_tcp_congestion_control, 812 }, 813 { 814 .procname = "tcp_keepalive_time", 815 .data = &init_net.ipv4.sysctl_tcp_keepalive_time, 816 .maxlen = sizeof(int), 817 .mode = 0644, 818 .proc_handler = proc_dointvec_jiffies, 819 }, 820 { 821 .procname = "tcp_keepalive_probes", 822 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes, 823 .maxlen = sizeof(int), 824 .mode = 0644, 825 .proc_handler = proc_dointvec 826 }, 827 { 828 .procname = "tcp_keepalive_intvl", 829 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl, 830 .maxlen = sizeof(int), 831 .mode = 0644, 832 .proc_handler = proc_dointvec_jiffies, 833 }, 834 { 835 .procname = "tcp_syn_retries", 836 .data = &init_net.ipv4.sysctl_tcp_syn_retries, 837 .maxlen = sizeof(int), 838 .mode = 0644, 839 .proc_handler = proc_dointvec_minmax, 840 .extra1 = &tcp_syn_retries_min, 841 .extra2 = &tcp_syn_retries_max 842 }, 843 { 844 .procname = "tcp_synack_retries", 845 .data = &init_net.ipv4.sysctl_tcp_synack_retries, 846 .maxlen = sizeof(int), 847 .mode = 0644, 848 .proc_handler = proc_dointvec 849 }, 850 #ifdef CONFIG_SYN_COOKIES 851 { 852 .procname = "tcp_syncookies", 853 .data = &init_net.ipv4.sysctl_tcp_syncookies, 854 .maxlen = sizeof(int), 855 .mode = 0644, 856 .proc_handler = proc_dointvec 857 }, 858 #endif 859 { 860 .procname = "tcp_reordering", 861 .data = &init_net.ipv4.sysctl_tcp_reordering, 862 .maxlen = sizeof(int), 863 .mode = 0644, 864 .proc_handler = proc_dointvec 865 }, 866 { 867 .procname = "tcp_retries1", 868 .data = &init_net.ipv4.sysctl_tcp_retries1, 869 .maxlen = sizeof(int), 870 .mode = 0644, 871 .proc_handler = proc_dointvec_minmax, 872 .extra2 = &tcp_retr1_max 873 }, 874 { 875 .procname = "tcp_retries2", 876 .data = &init_net.ipv4.sysctl_tcp_retries2, 877 .maxlen = sizeof(int), 878 .mode = 0644, 879 .proc_handler = proc_dointvec 880 }, 881 { 882 .procname = "tcp_orphan_retries", 883 .data = &init_net.ipv4.sysctl_tcp_orphan_retries, 884 .maxlen = sizeof(int), 885 .mode = 0644, 886 .proc_handler = proc_dointvec 887 }, 888 { 889 .procname = "tcp_fin_timeout", 890 .data = &init_net.ipv4.sysctl_tcp_fin_timeout, 891 .maxlen = sizeof(int), 892 .mode = 0644, 893 .proc_handler = proc_dointvec_jiffies, 894 }, 895 { 896 .procname = "tcp_notsent_lowat", 897 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat, 898 .maxlen = sizeof(unsigned int), 899 .mode = 0644, 900 .proc_handler = proc_douintvec, 901 }, 902 { 903 .procname = "tcp_tw_reuse", 904 .data = &init_net.ipv4.sysctl_tcp_tw_reuse, 905 .maxlen = sizeof(int), 906 .mode = 0644, 907 .proc_handler = proc_dointvec_minmax, 908 .extra1 = &zero, 909 .extra2 = &two, 910 }, 911 { 912 .procname = "tcp_max_tw_buckets", 913 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets, 914 .maxlen = sizeof(int), 915 .mode = 0644, 916 .proc_handler = proc_dointvec 917 }, 918 { 919 .procname = "tcp_max_syn_backlog", 920 .data = &init_net.ipv4.sysctl_max_syn_backlog, 921 .maxlen = sizeof(int), 922 .mode = 0644, 923 .proc_handler = proc_dointvec 924 }, 925 { 926 .procname = "tcp_fastopen", 927 .data = &init_net.ipv4.sysctl_tcp_fastopen, 928 .maxlen = sizeof(int), 929 .mode = 0644, 930 .proc_handler = proc_dointvec, 931 }, 932 { 933 .procname = "tcp_fastopen_key", 934 .mode = 0600, 935 .data = &init_net.ipv4.sysctl_tcp_fastopen, 936 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10), 937 .proc_handler = proc_tcp_fastopen_key, 938 }, 939 { 940 .procname = "tcp_fastopen_blackhole_timeout_sec", 941 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout, 942 .maxlen = sizeof(int), 943 .mode = 0644, 944 .proc_handler = proc_tfo_blackhole_detect_timeout, 945 .extra1 = &zero, 946 }, 947 #ifdef CONFIG_IP_ROUTE_MULTIPATH 948 { 949 .procname = "fib_multipath_use_neigh", 950 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh, 951 .maxlen = sizeof(int), 952 .mode = 0644, 953 .proc_handler = proc_dointvec_minmax, 954 .extra1 = &zero, 955 .extra2 = &one, 956 }, 957 { 958 .procname = "fib_multipath_hash_policy", 959 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy, 960 .maxlen = sizeof(int), 961 .mode = 0644, 962 .proc_handler = proc_fib_multipath_hash_policy, 963 .extra1 = &zero, 964 .extra2 = &one, 965 }, 966 #endif 967 { 968 .procname = "ip_unprivileged_port_start", 969 .maxlen = sizeof(int), 970 .data = &init_net.ipv4.sysctl_ip_prot_sock, 971 .mode = 0644, 972 .proc_handler = ipv4_privileged_ports, 973 }, 974 #ifdef CONFIG_NET_L3_MASTER_DEV 975 { 976 .procname = "udp_l3mdev_accept", 977 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept, 978 .maxlen = sizeof(int), 979 .mode = 0644, 980 .proc_handler = proc_dointvec_minmax, 981 .extra1 = &zero, 982 .extra2 = &one, 983 }, 984 #endif 985 { 986 .procname = "tcp_sack", 987 .data = &init_net.ipv4.sysctl_tcp_sack, 988 .maxlen = sizeof(int), 989 .mode = 0644, 990 .proc_handler = proc_dointvec 991 }, 992 { 993 .procname = "tcp_window_scaling", 994 .data = &init_net.ipv4.sysctl_tcp_window_scaling, 995 .maxlen = sizeof(int), 996 .mode = 0644, 997 .proc_handler = proc_dointvec 998 }, 999 { 1000 .procname = "tcp_timestamps", 1001 .data = &init_net.ipv4.sysctl_tcp_timestamps, 1002 .maxlen = sizeof(int), 1003 .mode = 0644, 1004 .proc_handler = proc_dointvec 1005 }, 1006 { 1007 .procname = "tcp_early_retrans", 1008 .data = &init_net.ipv4.sysctl_tcp_early_retrans, 1009 .maxlen = sizeof(int), 1010 .mode = 0644, 1011 .proc_handler = proc_dointvec_minmax, 1012 .extra1 = &zero, 1013 .extra2 = &four, 1014 }, 1015 { 1016 .procname = "tcp_recovery", 1017 .data = &init_net.ipv4.sysctl_tcp_recovery, 1018 .maxlen = sizeof(int), 1019 .mode = 0644, 1020 .proc_handler = proc_dointvec, 1021 }, 1022 { 1023 .procname = "tcp_thin_linear_timeouts", 1024 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts, 1025 .maxlen = sizeof(int), 1026 .mode = 0644, 1027 .proc_handler = proc_dointvec 1028 }, 1029 { 1030 .procname = "tcp_slow_start_after_idle", 1031 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle, 1032 .maxlen = sizeof(int), 1033 .mode = 0644, 1034 .proc_handler = proc_dointvec 1035 }, 1036 { 1037 .procname = "tcp_retrans_collapse", 1038 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse, 1039 .maxlen = sizeof(int), 1040 .mode = 0644, 1041 .proc_handler = proc_dointvec 1042 }, 1043 { 1044 .procname = "tcp_stdurg", 1045 .data = &init_net.ipv4.sysctl_tcp_stdurg, 1046 .maxlen = sizeof(int), 1047 .mode = 0644, 1048 .proc_handler = proc_dointvec 1049 }, 1050 { 1051 .procname = "tcp_rfc1337", 1052 .data = &init_net.ipv4.sysctl_tcp_rfc1337, 1053 .maxlen = sizeof(int), 1054 .mode = 0644, 1055 .proc_handler = proc_dointvec 1056 }, 1057 { 1058 .procname = "tcp_abort_on_overflow", 1059 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow, 1060 .maxlen = sizeof(int), 1061 .mode = 0644, 1062 .proc_handler = proc_dointvec 1063 }, 1064 { 1065 .procname = "tcp_fack", 1066 .data = &init_net.ipv4.sysctl_tcp_fack, 1067 .maxlen = sizeof(int), 1068 .mode = 0644, 1069 .proc_handler = proc_dointvec 1070 }, 1071 { 1072 .procname = "tcp_max_reordering", 1073 .data = &init_net.ipv4.sysctl_tcp_max_reordering, 1074 .maxlen = sizeof(int), 1075 .mode = 0644, 1076 .proc_handler = proc_dointvec 1077 }, 1078 { 1079 .procname = "tcp_dsack", 1080 .data = &init_net.ipv4.sysctl_tcp_dsack, 1081 .maxlen = sizeof(int), 1082 .mode = 0644, 1083 .proc_handler = proc_dointvec 1084 }, 1085 { 1086 .procname = "tcp_app_win", 1087 .data = &init_net.ipv4.sysctl_tcp_app_win, 1088 .maxlen = sizeof(int), 1089 .mode = 0644, 1090 .proc_handler = proc_dointvec 1091 }, 1092 { 1093 .procname = "tcp_adv_win_scale", 1094 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale, 1095 .maxlen = sizeof(int), 1096 .mode = 0644, 1097 .proc_handler = proc_dointvec_minmax, 1098 .extra1 = &tcp_adv_win_scale_min, 1099 .extra2 = &tcp_adv_win_scale_max, 1100 }, 1101 { 1102 .procname = "tcp_frto", 1103 .data = &init_net.ipv4.sysctl_tcp_frto, 1104 .maxlen = sizeof(int), 1105 .mode = 0644, 1106 .proc_handler = proc_dointvec 1107 }, 1108 { 1109 .procname = "tcp_no_metrics_save", 1110 .data = &init_net.ipv4.sysctl_tcp_nometrics_save, 1111 .maxlen = sizeof(int), 1112 .mode = 0644, 1113 .proc_handler = proc_dointvec, 1114 }, 1115 { 1116 .procname = "tcp_moderate_rcvbuf", 1117 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf, 1118 .maxlen = sizeof(int), 1119 .mode = 0644, 1120 .proc_handler = proc_dointvec, 1121 }, 1122 { 1123 .procname = "tcp_tso_win_divisor", 1124 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor, 1125 .maxlen = sizeof(int), 1126 .mode = 0644, 1127 .proc_handler = proc_dointvec, 1128 }, 1129 { 1130 .procname = "tcp_workaround_signed_windows", 1131 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows, 1132 .maxlen = sizeof(int), 1133 .mode = 0644, 1134 .proc_handler = proc_dointvec 1135 }, 1136 { 1137 .procname = "tcp_limit_output_bytes", 1138 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes, 1139 .maxlen = sizeof(int), 1140 .mode = 0644, 1141 .proc_handler = proc_dointvec 1142 }, 1143 { 1144 .procname = "tcp_challenge_ack_limit", 1145 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit, 1146 .maxlen = sizeof(int), 1147 .mode = 0644, 1148 .proc_handler = proc_dointvec 1149 }, 1150 { 1151 .procname = "tcp_min_tso_segs", 1152 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs, 1153 .maxlen = sizeof(int), 1154 .mode = 0644, 1155 .proc_handler = proc_dointvec_minmax, 1156 .extra1 = &one, 1157 .extra2 = &gso_max_segs, 1158 }, 1159 { 1160 .procname = "tcp_min_rtt_wlen", 1161 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen, 1162 .maxlen = sizeof(int), 1163 .mode = 0644, 1164 .proc_handler = proc_dointvec_minmax, 1165 .extra1 = &zero, 1166 .extra2 = &one_day_secs 1167 }, 1168 { 1169 .procname = "tcp_autocorking", 1170 .data = &init_net.ipv4.sysctl_tcp_autocorking, 1171 .maxlen = sizeof(int), 1172 .mode = 0644, 1173 .proc_handler = proc_dointvec_minmax, 1174 .extra1 = &zero, 1175 .extra2 = &one, 1176 }, 1177 { 1178 .procname = "tcp_invalid_ratelimit", 1179 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit, 1180 .maxlen = sizeof(int), 1181 .mode = 0644, 1182 .proc_handler = proc_dointvec_ms_jiffies, 1183 }, 1184 { 1185 .procname = "tcp_pacing_ss_ratio", 1186 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio, 1187 .maxlen = sizeof(int), 1188 .mode = 0644, 1189 .proc_handler = proc_dointvec_minmax, 1190 .extra1 = &zero, 1191 .extra2 = &thousand, 1192 }, 1193 { 1194 .procname = "tcp_pacing_ca_ratio", 1195 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio, 1196 .maxlen = sizeof(int), 1197 .mode = 0644, 1198 .proc_handler = proc_dointvec_minmax, 1199 .extra1 = &zero, 1200 .extra2 = &thousand, 1201 }, 1202 { 1203 .procname = "tcp_wmem", 1204 .data = &init_net.ipv4.sysctl_tcp_wmem, 1205 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem), 1206 .mode = 0644, 1207 .proc_handler = proc_dointvec_minmax, 1208 .extra1 = &one, 1209 }, 1210 { 1211 .procname = "tcp_rmem", 1212 .data = &init_net.ipv4.sysctl_tcp_rmem, 1213 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem), 1214 .mode = 0644, 1215 .proc_handler = proc_dointvec_minmax, 1216 .extra1 = &one, 1217 }, 1218 { 1219 .procname = "tcp_comp_sack_delay_ns", 1220 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns, 1221 .maxlen = sizeof(unsigned long), 1222 .mode = 0644, 1223 .proc_handler = proc_doulongvec_minmax, 1224 }, 1225 { 1226 .procname = "tcp_comp_sack_nr", 1227 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr, 1228 .maxlen = sizeof(int), 1229 .mode = 0644, 1230 .proc_handler = proc_dointvec_minmax, 1231 .extra1 = &zero, 1232 .extra2 = &comp_sack_nr_max, 1233 }, 1234 { 1235 .procname = "udp_rmem_min", 1236 .data = &init_net.ipv4.sysctl_udp_rmem_min, 1237 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min), 1238 .mode = 0644, 1239 .proc_handler = proc_dointvec_minmax, 1240 .extra1 = &one 1241 }, 1242 { 1243 .procname = "udp_wmem_min", 1244 .data = &init_net.ipv4.sysctl_udp_wmem_min, 1245 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min), 1246 .mode = 0644, 1247 .proc_handler = proc_dointvec_minmax, 1248 .extra1 = &one 1249 }, 1250 { } 1251 }; 1252 1253 static __net_init int ipv4_sysctl_init_net(struct net *net) 1254 { 1255 struct ctl_table *table; 1256 1257 table = ipv4_net_table; 1258 if (!net_eq(net, &init_net)) { 1259 int i; 1260 1261 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL); 1262 if (!table) 1263 goto err_alloc; 1264 1265 /* Update the variables to point into the current struct net */ 1266 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) 1267 table[i].data += (void *)net - (void *)&init_net; 1268 } 1269 1270 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table); 1271 if (!net->ipv4.ipv4_hdr) 1272 goto err_reg; 1273 1274 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL); 1275 if (!net->ipv4.sysctl_local_reserved_ports) 1276 goto err_ports; 1277 1278 return 0; 1279 1280 err_ports: 1281 unregister_net_sysctl_table(net->ipv4.ipv4_hdr); 1282 err_reg: 1283 if (!net_eq(net, &init_net)) 1284 kfree(table); 1285 err_alloc: 1286 return -ENOMEM; 1287 } 1288 1289 static __net_exit void ipv4_sysctl_exit_net(struct net *net) 1290 { 1291 struct ctl_table *table; 1292 1293 kfree(net->ipv4.sysctl_local_reserved_ports); 1294 table = net->ipv4.ipv4_hdr->ctl_table_arg; 1295 unregister_net_sysctl_table(net->ipv4.ipv4_hdr); 1296 kfree(table); 1297 } 1298 1299 static __net_initdata struct pernet_operations ipv4_sysctl_ops = { 1300 .init = ipv4_sysctl_init_net, 1301 .exit = ipv4_sysctl_exit_net, 1302 }; 1303 1304 static __init int sysctl_ipv4_init(void) 1305 { 1306 struct ctl_table_header *hdr; 1307 1308 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table); 1309 if (!hdr) 1310 return -ENOMEM; 1311 1312 if (register_pernet_subsys(&ipv4_sysctl_ops)) { 1313 unregister_net_sysctl_table(hdr); 1314 return -ENOMEM; 1315 } 1316 1317 return 0; 1318 } 1319 1320 __initcall(sysctl_ipv4_init); 1321