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