1.. SPDX-License-Identifier: GPL-2.0 2 3========= 4IP Sysctl 5========= 6 7/proc/sys/net/ipv4/* Variables 8============================== 9 10ip_forward - BOOLEAN 11 Forward Packets between interfaces. 12 13 This variable is special, its change resets all configuration 14 parameters to their default state (RFC1122 for hosts, RFC1812 15 for routers) 16 17 Possible values: 18 19 - 0 (disabled) 20 - 1 (enabled) 21 22 Default: 0 (disabled) 23 24ip_default_ttl - INTEGER 25 Default value of TTL field (Time To Live) for outgoing (but not 26 forwarded) IP packets. Should be between 1 and 255 inclusive. 27 Default: 64 (as recommended by RFC1700) 28 29ip_no_pmtu_disc - INTEGER 30 Disable Path MTU Discovery. If enabled in mode 1 and a 31 fragmentation-required ICMP is received, the PMTU to this 32 destination will be set to the smallest of the old MTU to 33 this destination and min_pmtu (see below). You will need 34 to raise min_pmtu to the smallest interface MTU on your system 35 manually if you want to avoid locally generated fragments. 36 37 In mode 2 incoming Path MTU Discovery messages will be 38 discarded. Outgoing frames are handled the same as in mode 1, 39 implicitly setting IP_PMTUDISC_DONT on every created socket. 40 41 Mode 3 is a hardened pmtu discover mode. The kernel will only 42 accept fragmentation-needed errors if the underlying protocol 43 can verify them besides a plain socket lookup. Current 44 protocols for which pmtu events will be honored are TCP and 45 SCTP as they verify e.g. the sequence number or the 46 association. This mode should not be enabled globally but is 47 only intended to secure e.g. name servers in namespaces where 48 TCP path mtu must still work but path MTU information of other 49 protocols should be discarded. If enabled globally this mode 50 could break other protocols. 51 52 Possible values: 0-3 53 54 Default: FALSE 55 56min_pmtu - INTEGER 57 default 552 - minimum Path MTU. Unless this is changed manually, 58 each cached pmtu will never be lower than this setting. 59 60ip_forward_use_pmtu - BOOLEAN 61 By default we don't trust protocol path MTUs while forwarding 62 because they could be easily forged and can lead to unwanted 63 fragmentation by the router. 64 You only need to enable this if you have user-space software 65 which tries to discover path mtus by itself and depends on the 66 kernel honoring this information. This is normally not the 67 case. 68 69 Possible values: 70 71 - 0 (disabled) 72 - 1 (enabled) 73 74 Default: 0 (disabled) 75 76fwmark_reflect - BOOLEAN 77 Controls the fwmark of kernel-generated IPv4 reply packets that are not 78 associated with a socket for example, TCP RSTs or ICMP echo replies). 79 If disabled, these packets have a fwmark of zero. If enabled, they have the 80 fwmark of the packet they are replying to. 81 82 Possible values: 83 84 - 0 (disabled) 85 - 1 (enabled) 86 87 Default: 0 (disabled) 88 89fib_multipath_use_neigh - BOOLEAN 90 Use status of existing neighbor entry when determining nexthop for 91 multipath routes. If disabled, neighbor information is not used and 92 packets could be directed to a failed nexthop. Only valid for kernels 93 built with CONFIG_IP_ROUTE_MULTIPATH enabled. 94 95 Possible values: 96 97 - 0 (disabled) 98 - 1 (enabled) 99 100 Default: 0 (disabled) 101 102fib_multipath_hash_policy - INTEGER 103 Controls which hash policy to use for multipath routes. Only valid 104 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled. 105 106 Default: 0 (Layer 3) 107 108 Possible values: 109 110 - 0 - Layer 3 111 - 1 - Layer 4 112 - 2 - Layer 3 or inner Layer 3 if present 113 - 3 - Custom multipath hash. Fields used for multipath hash calculation 114 are determined by fib_multipath_hash_fields sysctl 115 116fib_multipath_hash_fields - UNSIGNED INTEGER 117 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the 118 fields used for multipath hash calculation are determined by this 119 sysctl. 120 121 This value is a bitmask which enables various fields for multipath hash 122 calculation. 123 124 Possible fields are: 125 126 ====== ============================ 127 0x0001 Source IP address 128 0x0002 Destination IP address 129 0x0004 IP protocol 130 0x0008 Unused (Flow Label) 131 0x0010 Source port 132 0x0020 Destination port 133 0x0040 Inner source IP address 134 0x0080 Inner destination IP address 135 0x0100 Inner IP protocol 136 0x0200 Inner Flow Label 137 0x0400 Inner source port 138 0x0800 Inner destination port 139 ====== ============================ 140 141 Default: 0x0007 (source IP, destination IP and IP protocol) 142 143fib_multipath_hash_seed - UNSIGNED INTEGER 144 The seed value used when calculating hash for multipath routes. Applies 145 to both IPv4 and IPv6 datapath. Only present for kernels built with 146 CONFIG_IP_ROUTE_MULTIPATH enabled. 147 148 When set to 0, the seed value used for multipath routing defaults to an 149 internal random-generated one. 150 151 The actual hashing algorithm is not specified -- there is no guarantee 152 that a next hop distribution effected by a given seed will keep stable 153 across kernel versions. 154 155 Default: 0 (random) 156 157fib_sync_mem - UNSIGNED INTEGER 158 Amount of dirty memory from fib entries that can be backlogged before 159 synchronize_rcu is forced. 160 161 Default: 512kB Minimum: 64kB Maximum: 64MB 162 163ip_forward_update_priority - INTEGER 164 Whether to update SKB priority from "TOS" field in IPv4 header after it 165 is forwarded. The new SKB priority is mapped from TOS field value 166 according to an rt_tos2priority table (see e.g. man tc-prio). 167 168 Default: 1 (Update priority.) 169 170 Possible values: 171 172 - 0 - Do not update priority. 173 - 1 - Update priority. 174 175route/max_size - INTEGER 176 Maximum number of routes allowed in the kernel. Increase 177 this when using large numbers of interfaces and/or routes. 178 179 From linux kernel 3.6 onwards, this is deprecated for ipv4 180 as route cache is no longer used. 181 182 From linux kernel 6.3 onwards, this is deprecated for ipv6 183 as garbage collection manages cached route entries. 184 185neigh/default/gc_thresh1 - INTEGER 186 Minimum number of entries to keep. Garbage collector will not 187 purge entries if there are fewer than this number. 188 189 Default: 128 190 191neigh/default/gc_thresh2 - INTEGER 192 Threshold when garbage collector becomes more aggressive about 193 purging entries. Entries older than 5 seconds will be cleared 194 when over this number. 195 196 Default: 512 197 198neigh/default/gc_thresh3 - INTEGER 199 Maximum number of non-PERMANENT neighbor entries allowed. Increase 200 this when using large numbers of interfaces and when communicating 201 with large numbers of directly-connected peers. 202 203 Default: 1024 204 205neigh/default/unres_qlen_bytes - INTEGER 206 The maximum number of bytes which may be used by packets 207 queued for each unresolved address by other network layers. 208 (added in linux 3.3) 209 210 Setting negative value is meaningless and will return error. 211 212 Default: SK_WMEM_MAX, (same as net.core.wmem_default). 213 214 Exact value depends on architecture and kernel options, 215 but should be enough to allow queuing 256 packets 216 of medium size. 217 218neigh/default/unres_qlen - INTEGER 219 The maximum number of packets which may be queued for each 220 unresolved address by other network layers. 221 222 (deprecated in linux 3.3) : use unres_qlen_bytes instead. 223 224 Prior to linux 3.3, the default value is 3 which may cause 225 unexpected packet loss. The current default value is calculated 226 according to default value of unres_qlen_bytes and true size of 227 packet. 228 229 Default: 101 230 231neigh/default/interval_probe_time_ms - INTEGER 232 The probe interval for neighbor entries with NTF_MANAGED flag, 233 the min value is 1. 234 235 Default: 5000 236 237mtu_expires - INTEGER 238 Time, in seconds, that cached PMTU information is kept. 239 240min_adv_mss - INTEGER 241 The advertised MSS depends on the first hop route MTU, but will 242 never be lower than this setting. 243 244fib_notify_on_flag_change - INTEGER 245 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/ 246 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed. 247 248 After installing a route to the kernel, user space receives an 249 acknowledgment, which means the route was installed in the kernel, 250 but not necessarily in hardware. 251 It is also possible for a route already installed in hardware to change 252 its action and therefore its flags. For example, a host route that is 253 trapping packets can be "promoted" to perform decapsulation following 254 the installation of an IPinIP/VXLAN tunnel. 255 The notifications will indicate to user-space the state of the route. 256 257 Default: 0 (Do not emit notifications.) 258 259 Possible values: 260 261 - 0 - Do not emit notifications. 262 - 1 - Emit notifications. 263 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change. 264 265IP Fragmentation: 266 267ipfrag_high_thresh - LONG INTEGER 268 Maximum memory used to reassemble IP fragments. 269 270ipfrag_low_thresh - LONG INTEGER 271 (Obsolete since linux-4.17) 272 Maximum memory used to reassemble IP fragments before the kernel 273 begins to remove incomplete fragment queues to free up resources. 274 The kernel still accepts new fragments for defragmentation. 275 276ipfrag_time - INTEGER 277 Time in seconds to keep an IP fragment in memory. 278 279ipfrag_max_dist - INTEGER 280 ipfrag_max_dist is a non-negative integer value which defines the 281 maximum "disorder" which is allowed among fragments which share a 282 common IP source address. Note that reordering of packets is 283 not unusual, but if a large number of fragments arrive from a source 284 IP address while a particular fragment queue remains incomplete, it 285 probably indicates that one or more fragments belonging to that queue 286 have been lost. When ipfrag_max_dist is positive, an additional check 287 is done on fragments before they are added to a reassembly queue - if 288 ipfrag_max_dist (or more) fragments have arrived from a particular IP 289 address between additions to any IP fragment queue using that source 290 address, it's presumed that one or more fragments in the queue are 291 lost. The existing fragment queue will be dropped, and a new one 292 started. An ipfrag_max_dist value of zero disables this check. 293 294 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can 295 result in unnecessarily dropping fragment queues when normal 296 reordering of packets occurs, which could lead to poor application 297 performance. Using a very large value, e.g. 50000, increases the 298 likelihood of incorrectly reassembling IP fragments that originate 299 from different IP datagrams, which could result in data corruption. 300 Default: 64 301 302bc_forwarding - INTEGER 303 bc_forwarding enables the feature described in rfc1812#section-5.3.5.2 304 and rfc2644. It allows the router to forward directed broadcast. 305 To enable this feature, the 'all' entry and the input interface entry 306 should be set to 1. 307 Default: 0 308 309INET peer storage 310================= 311 312inet_peer_threshold - INTEGER 313 The approximate size of the storage. Starting from this threshold 314 entries will be thrown aggressively. This threshold also determines 315 entries' time-to-live and time intervals between garbage collection 316 passes. More entries, less time-to-live, less GC interval. 317 318inet_peer_minttl - INTEGER 319 Minimum time-to-live of entries. Should be enough to cover fragment 320 time-to-live on the reassembling side. This minimum time-to-live is 321 guaranteed if the pool size is less than inet_peer_threshold. 322 Measured in seconds. 323 324inet_peer_maxttl - INTEGER 325 Maximum time-to-live of entries. Unused entries will expire after 326 this period of time if there is no memory pressure on the pool (i.e. 327 when the number of entries in the pool is very small). 328 Measured in seconds. 329 330TCP variables 331============= 332 333somaxconn - INTEGER 334 Limit of socket listen() backlog, known in userspace as SOMAXCONN. 335 Defaults to 4096. (Was 128 before linux-5.4) 336 See also tcp_max_syn_backlog for additional tuning for TCP sockets. 337 338tcp_abort_on_overflow - BOOLEAN 339 If listening service is too slow to accept new connections, 340 reset them. Default state is FALSE. It means that if overflow 341 occurred due to a burst, connection will recover. Enable this 342 option _only_ if you are really sure that listening daemon 343 cannot be tuned to accept connections faster. Enabling this 344 option can harm clients of your server. 345 346tcp_adv_win_scale - INTEGER 347 Obsolete since linux-6.6 348 Count buffering overhead as bytes/2^tcp_adv_win_scale 349 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale), 350 if it is <= 0. 351 352 Possible values are [-31, 31], inclusive. 353 354 Default: 1 355 356tcp_allowed_congestion_control - STRING 357 Show/set the congestion control choices available to non-privileged 358 processes. The list is a subset of those listed in 359 tcp_available_congestion_control. 360 361 Default is "reno" and the default setting (tcp_congestion_control). 362 363tcp_app_win - INTEGER 364 Reserve max(window/2^tcp_app_win, mss) of window for application 365 buffer. Value 0 is special, it means that nothing is reserved. 366 367 Possible values are [0, 31], inclusive. 368 369 Default: 31 370 371tcp_autocorking - BOOLEAN 372 Enable TCP auto corking : 373 When applications do consecutive small write()/sendmsg() system calls, 374 we try to coalesce these small writes as much as possible, to lower 375 total amount of sent packets. This is done if at least one prior 376 packet for the flow is waiting in Qdisc queues or device transmit 377 queue. Applications can still use TCP_CORK for optimal behavior 378 when they know how/when to uncork their sockets. 379 380 Possible values: 381 382 - 0 (disabled) 383 - 1 (enabled) 384 385 Default: 1 (enabled) 386 387tcp_available_congestion_control - STRING 388 Shows the available congestion control choices that are registered. 389 More congestion control algorithms may be available as modules, 390 but not loaded. 391 392tcp_base_mss - INTEGER 393 The initial value of search_low to be used by the packetization layer 394 Path MTU discovery (MTU probing). If MTU probing is enabled, 395 this is the initial MSS used by the connection. 396 397tcp_mtu_probe_floor - INTEGER 398 If MTU probing is enabled this caps the minimum MSS used for search_low 399 for the connection. 400 401 Default : 48 402 403tcp_min_snd_mss - INTEGER 404 TCP SYN and SYNACK messages usually advertise an ADVMSS option, 405 as described in RFC 1122 and RFC 6691. 406 407 If this ADVMSS option is smaller than tcp_min_snd_mss, 408 it is silently capped to tcp_min_snd_mss. 409 410 Default : 48 (at least 8 bytes of payload per segment) 411 412tcp_congestion_control - STRING 413 Set the congestion control algorithm to be used for new 414 connections. The algorithm "reno" is always available, but 415 additional choices may be available based on kernel configuration. 416 Default is set as part of kernel configuration. 417 For passive connections, the listener congestion control choice 418 is inherited. 419 420 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ] 421 422tcp_dsack - BOOLEAN 423 Allows TCP to send "duplicate" SACKs. 424 425 Possible values: 426 427 - 0 (disabled) 428 - 1 (enabled) 429 430 Default: 1 (enabled) 431 432tcp_early_retrans - INTEGER 433 Tail loss probe (TLP) converts RTOs occurring due to tail 434 losses into fast recovery (RFC8985). Note that 435 TLP requires RACK to function properly (see tcp_recovery below) 436 437 Possible values: 438 439 - 0 disables TLP 440 - 3 or 4 enables TLP 441 442 Default: 3 443 444tcp_ecn - INTEGER 445 Control use of Explicit Congestion Notification (ECN) by TCP. 446 ECN is used only when both ends of the TCP connection indicate 447 support for it. This feature is useful in avoiding losses due 448 to congestion by allowing supporting routers to signal 449 congestion before having to drop packets. 450 451 Possible values are: 452 453 = ===================================================== 454 0 Disable ECN. Neither initiate nor accept ECN. 455 1 Enable ECN when requested by incoming connections and 456 also request ECN on outgoing connection attempts. 457 2 Enable ECN when requested by incoming connections 458 but do not request ECN on outgoing connections. 459 = ===================================================== 460 461 Default: 2 462 463tcp_ecn_fallback - BOOLEAN 464 If the kernel detects that ECN connection misbehaves, enable fall 465 back to non-ECN. Currently, this knob implements the fallback 466 from RFC3168, section 6.1.1.1., but we reserve that in future, 467 additional detection mechanisms could be implemented under this 468 knob. The value is not used, if tcp_ecn or per route (or congestion 469 control) ECN settings are disabled. 470 471 Possible values: 472 473 - 0 (disabled) 474 - 1 (enabled) 475 476 Default: 1 (enabled) 477 478tcp_fack - BOOLEAN 479 This is a legacy option, it has no effect anymore. 480 481tcp_fin_timeout - INTEGER 482 The length of time an orphaned (no longer referenced by any 483 application) connection will remain in the FIN_WAIT_2 state 484 before it is aborted at the local end. While a perfectly 485 valid "receive only" state for an un-orphaned connection, an 486 orphaned connection in FIN_WAIT_2 state could otherwise wait 487 forever for the remote to close its end of the connection. 488 489 Cf. tcp_max_orphans 490 491 Default: 60 seconds 492 493tcp_frto - INTEGER 494 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682. 495 F-RTO is an enhanced recovery algorithm for TCP retransmission 496 timeouts. It is particularly beneficial in networks where the 497 RTT fluctuates (e.g., wireless). F-RTO is sender-side only 498 modification. It does not require any support from the peer. 499 500 By default it's enabled with a non-zero value. 0 disables F-RTO. 501 502tcp_fwmark_accept - BOOLEAN 503 If enabled, incoming connections to listening sockets that do not have a 504 socket mark will set the mark of the accepting socket to the fwmark of 505 the incoming SYN packet. This will cause all packets on that connection 506 (starting from the first SYNACK) to be sent with that fwmark. The 507 listening socket's mark is unchanged. Listening sockets that already 508 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are 509 unaffected. 510 511 Possible values: 512 513 - 0 (disabled) 514 - 1 (enabled) 515 516 Default: 0 (disabled) 517 518tcp_invalid_ratelimit - INTEGER 519 Limit the maximal rate for sending duplicate acknowledgments 520 in response to incoming TCP packets that are for an existing 521 connection but that are invalid due to any of these reasons: 522 523 (a) out-of-window sequence number, 524 (b) out-of-window acknowledgment number, or 525 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure 526 527 This can help mitigate simple "ack loop" DoS attacks, wherein 528 a buggy or malicious middlebox or man-in-the-middle can 529 rewrite TCP header fields in manner that causes each endpoint 530 to think that the other is sending invalid TCP segments, thus 531 causing each side to send an unterminating stream of duplicate 532 acknowledgments for invalid segments. 533 534 Using 0 disables rate-limiting of dupacks in response to 535 invalid segments; otherwise this value specifies the minimal 536 space between sending such dupacks, in milliseconds. 537 538 Default: 500 (milliseconds). 539 540tcp_keepalive_time - INTEGER 541 How often TCP sends out keepalive messages when keepalive is enabled. 542 Default: 2hours. 543 544tcp_keepalive_probes - INTEGER 545 How many keepalive probes TCP sends out, until it decides that the 546 connection is broken. Default value: 9. 547 548tcp_keepalive_intvl - INTEGER 549 How frequently the probes are send out. Multiplied by 550 tcp_keepalive_probes it is time to kill not responding connection, 551 after probes started. Default value: 75sec i.e. connection 552 will be aborted after ~11 minutes of retries. 553 554tcp_l3mdev_accept - BOOLEAN 555 Enables child sockets to inherit the L3 master device index. 556 Enabling this option allows a "global" listen socket to work 557 across L3 master domains (e.g., VRFs) with connected sockets 558 derived from the listen socket to be bound to the L3 domain in 559 which the packets originated. Only valid when the kernel was 560 compiled with CONFIG_NET_L3_MASTER_DEV. 561 562 Possible values: 563 564 - 0 (disabled) 565 - 1 (enabled) 566 567 Default: 0 (disabled) 568 569tcp_low_latency - BOOLEAN 570 This is a legacy option, it has no effect anymore. 571 572tcp_max_orphans - INTEGER 573 Maximal number of TCP sockets not attached to any user file handle, 574 held by system. If this number is exceeded orphaned connections are 575 reset immediately and warning is printed. This limit exists 576 only to prevent simple DoS attacks, you _must_ not rely on this 577 or lower the limit artificially, but rather increase it 578 (probably, after increasing installed memory), 579 if network conditions require more than default value, 580 and tune network services to linger and kill such states 581 more aggressively. Let me to remind again: each orphan eats 582 up to ~64K of unswappable memory. 583 584tcp_max_syn_backlog - INTEGER 585 Maximal number of remembered connection requests (SYN_RECV), 586 which have not received an acknowledgment from connecting client. 587 588 This is a per-listener limit. 589 590 The minimal value is 128 for low memory machines, and it will 591 increase in proportion to the memory of machine. 592 593 If server suffers from overload, try increasing this number. 594 595 Remember to also check /proc/sys/net/core/somaxconn 596 A SYN_RECV request socket consumes about 304 bytes of memory. 597 598tcp_max_tw_buckets - INTEGER 599 Maximal number of timewait sockets held by system simultaneously. 600 If this number is exceeded time-wait socket is immediately destroyed 601 and warning is printed. This limit exists only to prevent 602 simple DoS attacks, you _must_ not lower the limit artificially, 603 but rather increase it (probably, after increasing installed memory), 604 if network conditions require more than default value. 605 606tcp_mem - vector of 3 INTEGERs: min, pressure, max 607 min: below this number of pages TCP is not bothered about its 608 memory appetite. 609 610 pressure: when amount of memory allocated by TCP exceeds this number 611 of pages, TCP moderates its memory consumption and enters memory 612 pressure mode, which is exited when memory consumption falls 613 under "min". 614 615 max: number of pages allowed for queueing by all TCP sockets. 616 617 Defaults are calculated at boot time from amount of available 618 memory. 619 620tcp_min_rtt_wlen - INTEGER 621 The window length of the windowed min filter to track the minimum RTT. 622 A shorter window lets a flow more quickly pick up new (higher) 623 minimum RTT when it is moved to a longer path (e.g., due to traffic 624 engineering). A longer window makes the filter more resistant to RTT 625 inflations such as transient congestion. The unit is seconds. 626 627 Possible values: 0 - 86400 (1 day) 628 629 Default: 300 630 631tcp_moderate_rcvbuf - BOOLEAN 632 If enabled, TCP performs receive buffer auto-tuning, attempting to 633 automatically size the buffer (no greater than tcp_rmem[2]) to 634 match the size required by the path for full throughput. 635 636 Possible values: 637 638 - 0 (disabled) 639 - 1 (enabled) 640 641 Default: 1 (enabled) 642 643tcp_mtu_probing - INTEGER 644 Controls TCP Packetization-Layer Path MTU Discovery. Takes three 645 values: 646 647 - 0 - Disabled 648 - 1 - Disabled by default, enabled when an ICMP black hole detected 649 - 2 - Always enabled, use initial MSS of tcp_base_mss. 650 651tcp_probe_interval - UNSIGNED INTEGER 652 Controls how often to start TCP Packetization-Layer Path MTU 653 Discovery reprobe. The default is reprobing every 10 minutes as 654 per RFC4821. 655 656tcp_probe_threshold - INTEGER 657 Controls when TCP Packetization-Layer Path MTU Discovery probing 658 will stop in respect to the width of search range in bytes. Default 659 is 8 bytes. 660 661tcp_no_metrics_save - BOOLEAN 662 By default, TCP saves various connection metrics in the route cache 663 when the connection closes, so that connections established in the 664 near future can use these to set initial conditions. Usually, this 665 increases overall performance, but may sometimes cause performance 666 degradation. If enabled, TCP will not cache metrics on closing 667 connections. 668 669 Possible values: 670 671 - 0 (disabled) 672 - 1 (enabled) 673 674 Default: 0 (disabled) 675 676tcp_no_ssthresh_metrics_save - BOOLEAN 677 Controls whether TCP saves ssthresh metrics in the route cache. 678 If enabled, ssthresh metrics are disabled. 679 680 Possible values: 681 682 - 0 (disabled) 683 - 1 (enabled) 684 685 Default: 1 (enabled) 686 687tcp_orphan_retries - INTEGER 688 This value influences the timeout of a locally closed TCP connection, 689 when RTO retransmissions remain unacknowledged. 690 See tcp_retries2 for more details. 691 692 The default value is 8. 693 694 If your machine is a loaded WEB server, 695 you should think about lowering this value, such sockets 696 may consume significant resources. Cf. tcp_max_orphans. 697 698tcp_recovery - INTEGER 699 This value is a bitmap to enable various experimental loss recovery 700 features. 701 702 ========= ============================================================= 703 RACK: 0x1 enables RACK loss detection, for fast detection of lost 704 retransmissions and tail drops, and resilience to 705 reordering. currently, setting this bit to 0 has no 706 effect, since RACK is the only supported loss detection 707 algorithm. 708 709 RACK: 0x2 makes RACK's reordering window static (min_rtt/4). 710 711 RACK: 0x4 disables RACK's DUPACK threshold heuristic 712 ========= ============================================================= 713 714 Default: 0x1 715 716tcp_reflect_tos - BOOLEAN 717 For listening sockets, reuse the DSCP value of the initial SYN message 718 for outgoing packets. This allows to have both directions of a TCP 719 stream to use the same DSCP value, assuming DSCP remains unchanged for 720 the lifetime of the connection. 721 722 This options affects both IPv4 and IPv6. 723 724 Possible values: 725 726 - 0 (disabled) 727 - 1 (enabled) 728 729 Default: 0 (disabled) 730 731tcp_reordering - INTEGER 732 Initial reordering level of packets in a TCP stream. 733 TCP stack can then dynamically adjust flow reordering level 734 between this initial value and tcp_max_reordering 735 736 Default: 3 737 738tcp_max_reordering - INTEGER 739 Maximal reordering level of packets in a TCP stream. 740 300 is a fairly conservative value, but you might increase it 741 if paths are using per packet load balancing (like bonding rr mode) 742 743 Default: 300 744 745tcp_retrans_collapse - BOOLEAN 746 Bug-to-bug compatibility with some broken printers. 747 On retransmit try to send bigger packets to work around bugs in 748 certain TCP stacks. 749 750 Possible values: 751 752 - 0 (disabled) 753 - 1 (enabled) 754 755 Default: 1 (enabled) 756 757tcp_retries1 - INTEGER 758 This value influences the time, after which TCP decides, that 759 something is wrong due to unacknowledged RTO retransmissions, 760 and reports this suspicion to the network layer. 761 See tcp_retries2 for more details. 762 763 RFC 1122 recommends at least 3 retransmissions, which is the 764 default. 765 766tcp_retries2 - INTEGER 767 This value influences the timeout of an alive TCP connection, 768 when RTO retransmissions remain unacknowledged. 769 Given a value of N, a hypothetical TCP connection following 770 exponential backoff with an initial RTO of TCP_RTO_MIN would 771 retransmit N times before killing the connection at the (N+1)th RTO. 772 773 The default value of 15 yields a hypothetical timeout of 924.6 774 seconds and is a lower bound for the effective timeout. 775 TCP will effectively time out at the first RTO which exceeds the 776 hypothetical timeout. 777 If tcp_rto_max_ms is decreased, it is recommended to also 778 change tcp_retries2. 779 780 RFC 1122 recommends at least 100 seconds for the timeout, 781 which corresponds to a value of at least 8. 782 783tcp_rfc1337 - BOOLEAN 784 If enabled, the TCP stack behaves conforming to RFC1337. If unset, 785 we are not conforming to RFC, but prevent TCP TIME_WAIT 786 assassination. 787 788 Possible values: 789 790 - 0 (disabled) 791 - 1 (enabled) 792 793 Default: 0 (disabled) 794 795tcp_rmem - vector of 3 INTEGERs: min, default, max 796 min: Minimal size of receive buffer used by TCP sockets. 797 It is guaranteed to each TCP socket, even under moderate memory 798 pressure. 799 800 Default: 4K 801 802 default: initial size of receive buffer used by TCP sockets. 803 This value overrides net.core.rmem_default used by other protocols. 804 Default: 131072 bytes. 805 This value results in initial window of 65535. 806 807 max: maximal size of receive buffer allowed for automatically 808 selected receiver buffers for TCP socket. This value does not override 809 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables 810 automatic tuning of that socket's receive buffer size, in which 811 case this value is ignored. 812 Default: between 131072 and 32MB, depending on RAM size. 813 814tcp_sack - BOOLEAN 815 Enable select acknowledgments (SACKS). 816 817 Possible values: 818 819 - 0 (disabled) 820 - 1 (enabled) 821 822 Default: 1 (enabled) 823 824tcp_comp_sack_delay_ns - LONG INTEGER 825 TCP tries to reduce number of SACK sent, using a timer 826 based on 5% of SRTT, capped by this sysctl, in nano seconds. 827 The default is 1ms, based on TSO autosizing period. 828 829 Default : 1,000,000 ns (1 ms) 830 831tcp_comp_sack_slack_ns - LONG INTEGER 832 This sysctl control the slack used when arming the 833 timer used by SACK compression. This gives extra time 834 for small RTT flows, and reduces system overhead by allowing 835 opportunistic reduction of timer interrupts. 836 837 Default : 100,000 ns (100 us) 838 839tcp_comp_sack_nr - INTEGER 840 Max number of SACK that can be compressed. 841 Using 0 disables SACK compression. 842 843 Default : 44 844 845tcp_backlog_ack_defer - BOOLEAN 846 If enabled, user thread processing socket backlog tries sending 847 one ACK for the whole queue. This helps to avoid potential 848 long latencies at end of a TCP socket syscall. 849 850 Possible values: 851 852 - 0 (disabled) 853 - 1 (enabled) 854 855 Default: 1 (enabled) 856 857tcp_slow_start_after_idle - BOOLEAN 858 If enabled, provide RFC2861 behavior and time out the congestion 859 window after an idle period. An idle period is defined at 860 the current RTO. If unset, the congestion window will not 861 be timed out after an idle period. 862 863 Possible values: 864 865 - 0 (disabled) 866 - 1 (enabled) 867 868 Default: 1 (enabled) 869 870tcp_stdurg - BOOLEAN 871 Use the Host requirements interpretation of the TCP urgent pointer field. 872 Most hosts use the older BSD interpretation, so if enabled, 873 Linux might not communicate correctly with them. 874 875 Possible values: 876 877 - 0 (disabled) 878 - 1 (enabled) 879 880 Default: 0 (disabled) 881 882tcp_synack_retries - INTEGER 883 Number of times SYNACKs for a passive TCP connection attempt will 884 be retransmitted. Should not be higher than 255. Default value 885 is 5, which corresponds to 31seconds till the last retransmission 886 with the current initial RTO of 1second. With this the final timeout 887 for a passive TCP connection will happen after 63seconds. 888 889tcp_syncookies - INTEGER 890 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES 891 Send out syncookies when the syn backlog queue of a socket 892 overflows. This is to prevent against the common 'SYN flood attack' 893 Default: 1 894 895 Note, that syncookies is fallback facility. 896 It MUST NOT be used to help highly loaded servers to stand 897 against legal connection rate. If you see SYN flood warnings 898 in your logs, but investigation shows that they occur 899 because of overload with legal connections, you should tune 900 another parameters until this warning disappear. 901 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow. 902 903 syncookies seriously violate TCP protocol, do not allow 904 to use TCP extensions, can result in serious degradation 905 of some services (f.e. SMTP relaying), visible not by you, 906 but your clients and relays, contacting you. While you see 907 SYN flood warnings in logs not being really flooded, your server 908 is seriously misconfigured. 909 910 If you want to test which effects syncookies have to your 911 network connections you can set this knob to 2 to enable 912 unconditionally generation of syncookies. 913 914tcp_migrate_req - BOOLEAN 915 The incoming connection is tied to a specific listening socket when 916 the initial SYN packet is received during the three-way handshake. 917 When a listener is closed, in-flight request sockets during the 918 handshake and established sockets in the accept queue are aborted. 919 920 If the listener has SO_REUSEPORT enabled, other listeners on the 921 same port should have been able to accept such connections. This 922 option makes it possible to migrate such child sockets to another 923 listener after close() or shutdown(). 924 925 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should 926 usually be used to define the policy to pick an alive listener. 927 Otherwise, the kernel will randomly pick an alive listener only if 928 this option is enabled. 929 930 Note that migration between listeners with different settings may 931 crash applications. Let's say migration happens from listener A to 932 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from 933 the requests migrated from A. To avoid such a situation, cancel 934 migration by returning SK_DROP in the type of eBPF program, or 935 disable this option. 936 937 Possible values: 938 939 - 0 (disabled) 940 - 1 (enabled) 941 942 Default: 0 (disabled) 943 944tcp_fastopen - INTEGER 945 Enable TCP Fast Open (RFC7413) to send and accept data in the opening 946 SYN packet. 947 948 The client support is enabled by flag 0x1 (on by default). The client 949 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag, 950 rather than connect() to send data in SYN. 951 952 The server support is enabled by flag 0x2 (off by default). Then 953 either enable for all listeners with another flag (0x400) or 954 enable individual listeners via TCP_FASTOPEN socket option with 955 the option value being the length of the syn-data backlog. 956 957 The values (bitmap) are 958 959 ===== ======== ====================================================== 960 0x1 (client) enables sending data in the opening SYN on the client. 961 0x2 (server) enables the server support, i.e., allowing data in 962 a SYN packet to be accepted and passed to the 963 application before 3-way handshake finishes. 964 0x4 (client) send data in the opening SYN regardless of cookie 965 availability and without a cookie option. 966 0x200 (server) accept data-in-SYN w/o any cookie option present. 967 0x400 (server) enable all listeners to support Fast Open by 968 default without explicit TCP_FASTOPEN socket option. 969 ===== ======== ====================================================== 970 971 Default: 0x1 972 973 Note that additional client or server features are only 974 effective if the basic support (0x1 and 0x2) are enabled respectively. 975 976tcp_fastopen_blackhole_timeout_sec - INTEGER 977 Initial time period in second to disable Fastopen on active TCP sockets 978 when a TFO firewall blackhole issue happens. 979 This time period will grow exponentially when more blackhole issues 980 get detected right after Fastopen is re-enabled and will reset to 981 initial value when the blackhole issue goes away. 982 0 to disable the blackhole detection. 983 984 By default, it is set to 0 (feature is disabled). 985 986tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs 987 The list consists of a primary key and an optional backup key. The 988 primary key is used for both creating and validating cookies, while the 989 optional backup key is only used for validating cookies. The purpose of 990 the backup key is to maximize TFO validation when keys are rotated. 991 992 A randomly chosen primary key may be configured by the kernel if 993 the tcp_fastopen sysctl is set to 0x400 (see above), or if the 994 TCP_FASTOPEN setsockopt() optname is set and a key has not been 995 previously configured via sysctl. If keys are configured via 996 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those 997 per-socket keys will be used instead of any keys that are specified via 998 sysctl. 999 1000 A key is specified as 4 8-digit hexadecimal integers which are separated 1001 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be 1002 omitted. A primary and a backup key may be specified by separating them 1003 by a comma. If only one key is specified, it becomes the primary key and 1004 any previously configured backup keys are removed. 1005 1006tcp_syn_retries - INTEGER 1007 Number of times initial SYNs for an active TCP connection attempt 1008 will be retransmitted. Should not be higher than 127. Default value 1009 is 6, which corresponds to 67seconds (with tcp_syn_linear_timeouts = 4) 1010 till the last retransmission with the current initial RTO of 1second. 1011 With this the final timeout for an active TCP connection attempt 1012 will happen after 131seconds. 1013 1014tcp_timestamps - INTEGER 1015 Enable timestamps as defined in RFC1323. 1016 1017 - 0: Disabled. 1018 - 1: Enable timestamps as defined in RFC1323 and use random offset for 1019 each connection rather than only using the current time. 1020 - 2: Like 1, but without random offsets. 1021 1022 Default: 1 1023 1024tcp_min_tso_segs - INTEGER 1025 Minimal number of segments per TSO frame. 1026 1027 Since linux-3.12, TCP does an automatic sizing of TSO frames, 1028 depending on flow rate, instead of filling 64Kbytes packets. 1029 For specific usages, it's possible to force TCP to build big 1030 TSO frames. Note that TCP stack might split too big TSO packets 1031 if available window is too small. 1032 1033 Default: 2 1034 1035tcp_tso_rtt_log - INTEGER 1036 Adjustment of TSO packet sizes based on min_rtt 1037 1038 Starting from linux-5.18, TCP autosizing can be tweaked 1039 for flows having small RTT. 1040 1041 Old autosizing was splitting the pacing budget to send 1024 TSO 1042 per second. 1043 1044 tso_packet_size = sk->sk_pacing_rate / 1024; 1045 1046 With the new mechanism, we increase this TSO sizing using: 1047 1048 distance = min_rtt_usec / (2^tcp_tso_rtt_log) 1049 tso_packet_size += gso_max_size >> distance; 1050 1051 This means that flows between very close hosts can use bigger 1052 TSO packets, reducing their cpu costs. 1053 1054 If you want to use the old autosizing, set this sysctl to 0. 1055 1056 Default: 9 (2^9 = 512 usec) 1057 1058tcp_pacing_ss_ratio - INTEGER 1059 sk->sk_pacing_rate is set by TCP stack using a ratio applied 1060 to current rate. (current_rate = cwnd * mss / srtt) 1061 If TCP is in slow start, tcp_pacing_ss_ratio is applied 1062 to let TCP probe for bigger speeds, assuming cwnd can be 1063 doubled every other RTT. 1064 1065 Default: 200 1066 1067tcp_pacing_ca_ratio - INTEGER 1068 sk->sk_pacing_rate is set by TCP stack using a ratio applied 1069 to current rate. (current_rate = cwnd * mss / srtt) 1070 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio 1071 is applied to conservatively probe for bigger throughput. 1072 1073 Default: 120 1074 1075tcp_syn_linear_timeouts - INTEGER 1076 The number of times for an active TCP connection to retransmit SYNs with 1077 a linear backoff timeout before defaulting to an exponential backoff 1078 timeout. This has no effect on SYNACK at the passive TCP side. 1079 1080 With an initial RTO of 1 and tcp_syn_linear_timeouts = 4 we would 1081 expect SYN RTOs to be: 1, 1, 1, 1, 1, 2, 4, ... (4 linear timeouts, 1082 and the first exponential backoff using 2^0 * initial_RTO). 1083 Default: 4 1084 1085tcp_tso_win_divisor - INTEGER 1086 This allows control over what percentage of the congestion window 1087 can be consumed by a single TSO frame. 1088 The setting of this parameter is a choice between burstiness and 1089 building larger TSO frames. 1090 1091 Default: 3 1092 1093tcp_tw_reuse - INTEGER 1094 Enable reuse of TIME-WAIT sockets for new connections when it is 1095 safe from protocol viewpoint. 1096 1097 - 0 - disable 1098 - 1 - global enable 1099 - 2 - enable for loopback traffic only 1100 1101 It should not be changed without advice/request of technical 1102 experts. 1103 1104 Default: 2 1105 1106tcp_tw_reuse_delay - UNSIGNED INTEGER 1107 The delay in milliseconds before a TIME-WAIT socket can be reused by a 1108 new connection, if TIME-WAIT socket reuse is enabled. The actual reuse 1109 threshold is within [N, N+1] range, where N is the requested delay in 1110 milliseconds, to ensure the delay interval is never shorter than the 1111 configured value. 1112 1113 This setting contains an assumption about the other TCP timestamp clock 1114 tick interval. It should not be set to a value lower than the peer's 1115 clock tick for PAWS (Protection Against Wrapped Sequence numbers) 1116 mechanism work correctly for the reused connection. 1117 1118 Default: 1000 (milliseconds) 1119 1120tcp_window_scaling - BOOLEAN 1121 Enable window scaling as defined in RFC1323. 1122 1123 Possible values: 1124 1125 - 0 (disabled) 1126 - 1 (enabled) 1127 1128 Default: 1 (enabled) 1129 1130tcp_shrink_window - BOOLEAN 1131 This changes how the TCP receive window is calculated. 1132 1133 RFC 7323, section 2.4, says there are instances when a retracted 1134 window can be offered, and that TCP implementations MUST ensure 1135 that they handle a shrinking window, as specified in RFC 1122. 1136 1137 Possible values: 1138 1139 - 0 (disabled) - The window is never shrunk. 1140 - 1 (enabled) - The window is shrunk when necessary to remain within 1141 the memory limit set by autotuning (sk_rcvbuf). 1142 This only occurs if a non-zero receive window 1143 scaling factor is also in effect. 1144 1145 Default: 0 (disabled) 1146 1147tcp_wmem - vector of 3 INTEGERs: min, default, max 1148 min: Amount of memory reserved for send buffers for TCP sockets. 1149 Each TCP socket has rights to use it due to fact of its birth. 1150 1151 Default: 4K 1152 1153 default: initial size of send buffer used by TCP sockets. This 1154 value overrides net.core.wmem_default used by other protocols. 1155 1156 It is usually lower than net.core.wmem_default. 1157 1158 Default: 16K 1159 1160 max: Maximal amount of memory allowed for automatically tuned 1161 send buffers for TCP sockets. This value does not override 1162 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables 1163 automatic tuning of that socket's send buffer size, in which case 1164 this value is ignored. 1165 1166 Default: between 64K and 4MB, depending on RAM size. 1167 1168tcp_notsent_lowat - UNSIGNED INTEGER 1169 A TCP socket can control the amount of unsent bytes in its write queue, 1170 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll() 1171 reports POLLOUT events if the amount of unsent bytes is below a per 1172 socket value, and if the write queue is not full. sendmsg() will 1173 also not add new buffers if the limit is hit. 1174 1175 This global variable controls the amount of unsent data for 1176 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change 1177 to the global variable has immediate effect. 1178 1179 Default: UINT_MAX (0xFFFFFFFF) 1180 1181tcp_workaround_signed_windows - BOOLEAN 1182 If enabled, assume no receipt of a window scaling option means the 1183 remote TCP is broken and treats the window as a signed quantity. 1184 If disabled, assume the remote TCP is not broken even if we do 1185 not receive a window scaling option from them. 1186 1187 Possible values: 1188 1189 - 0 (disabled) 1190 - 1 (enabled) 1191 1192 Default: 0 (disabled) 1193 1194tcp_thin_linear_timeouts - BOOLEAN 1195 Enable dynamic triggering of linear timeouts for thin streams. 1196 If enabled, a check is performed upon retransmission by timeout to 1197 determine if the stream is thin (less than 4 packets in flight). 1198 As long as the stream is found to be thin, up to 6 linear 1199 timeouts may be performed before exponential backoff mode is 1200 initiated. This improves retransmission latency for 1201 non-aggressive thin streams, often found to be time-dependent. 1202 For more information on thin streams, see 1203 Documentation/networking/tcp-thin.rst 1204 1205 Possible values: 1206 1207 - 0 (disabled) 1208 - 1 (enabled) 1209 1210 Default: 0 (disabled) 1211 1212tcp_limit_output_bytes - INTEGER 1213 Controls TCP Small Queue limit per tcp socket. 1214 TCP bulk sender tends to increase packets in flight until it 1215 gets losses notifications. With SNDBUF autotuning, this can 1216 result in a large amount of packets queued on the local machine 1217 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other 1218 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes 1219 limits the number of bytes on qdisc or device to reduce artificial 1220 RTT/cwnd and reduce bufferbloat. 1221 1222 Default: 4194304 (4 MB) 1223 1224tcp_challenge_ack_limit - INTEGER 1225 Limits number of Challenge ACK sent per second, as recommended 1226 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks) 1227 Note that this per netns rate limit can allow some side channel 1228 attacks and probably should not be enabled. 1229 TCP stack implements per TCP socket limits anyway. 1230 Default: INT_MAX (unlimited) 1231 1232tcp_ehash_entries - INTEGER 1233 Show the number of hash buckets for TCP sockets in the current 1234 networking namespace. 1235 1236 A negative value means the networking namespace does not own its 1237 hash buckets and shares the initial networking namespace's one. 1238 1239tcp_child_ehash_entries - INTEGER 1240 Control the number of hash buckets for TCP sockets in the child 1241 networking namespace, which must be set before clone() or unshare(). 1242 1243 If the value is not 0, the kernel uses a value rounded up to 2^n 1244 as the actual hash bucket size. 0 is a special value, meaning 1245 the child networking namespace will share the initial networking 1246 namespace's hash buckets. 1247 1248 Note that the child will use the global one in case the kernel 1249 fails to allocate enough memory. In addition, the global hash 1250 buckets are spread over available NUMA nodes, but the allocation 1251 of the child hash table depends on the current process's NUMA 1252 policy, which could result in performance differences. 1253 1254 Note also that the default value of tcp_max_tw_buckets and 1255 tcp_max_syn_backlog depend on the hash bucket size. 1256 1257 Possible values: 0, 2^n (n: 0 - 24 (16Mi)) 1258 1259 Default: 0 1260 1261tcp_plb_enabled - BOOLEAN 1262 If enabled and the underlying congestion control (e.g. DCTCP) supports 1263 and enables PLB feature, TCP PLB (Protective Load Balancing) is 1264 enabled. PLB is described in the following paper: 1265 https://doi.org/10.1145/3544216.3544226. Based on PLB parameters, 1266 upon sensing sustained congestion, TCP triggers a change in 1267 flow label field for outgoing IPv6 packets. A change in flow label 1268 field potentially changes the path of outgoing packets for switches 1269 that use ECMP/WCMP for routing. 1270 1271 PLB changes socket txhash which results in a change in IPv6 Flow Label 1272 field, and currently no-op for IPv4 headers. It is possible 1273 to apply PLB for IPv4 with other network header fields (e.g. TCP 1274 or IPv4 options) or using encapsulation where outer header is used 1275 by switches to determine next hop. In either case, further host 1276 and switch side changes will be needed. 1277 1278 If enabled, PLB assumes that congestion signal (e.g. ECN) is made 1279 available and used by congestion control module to estimate a 1280 congestion measure (e.g. ce_ratio). PLB needs a congestion measure to 1281 make repathing decisions. 1282 1283 Possible values: 1284 1285 - 0 (disabled) 1286 - 1 (enabled) 1287 1288 Default: 0 (disabled) 1289 1290tcp_plb_idle_rehash_rounds - INTEGER 1291 Number of consecutive congested rounds (RTT) seen after which 1292 a rehash can be performed, given there are no packets in flight. 1293 This is referred to as M in PLB paper: 1294 https://doi.org/10.1145/3544216.3544226. 1295 1296 Possible Values: 0 - 31 1297 1298 Default: 3 1299 1300tcp_plb_rehash_rounds - INTEGER 1301 Number of consecutive congested rounds (RTT) seen after which 1302 a forced rehash can be performed. Be careful when setting this 1303 parameter, as a small value increases the risk of retransmissions. 1304 This is referred to as N in PLB paper: 1305 https://doi.org/10.1145/3544216.3544226. 1306 1307 Possible Values: 0 - 31 1308 1309 Default: 12 1310 1311tcp_plb_suspend_rto_sec - INTEGER 1312 Time, in seconds, to suspend PLB in event of an RTO. In order to avoid 1313 having PLB repath onto a connectivity "black hole", after an RTO a TCP 1314 connection suspends PLB repathing for a random duration between 1x and 1315 2x of this parameter. Randomness is added to avoid concurrent rehashing 1316 of multiple TCP connections. This should be set corresponding to the 1317 amount of time it takes to repair a failed link. 1318 1319 Possible Values: 0 - 255 1320 1321 Default: 60 1322 1323tcp_plb_cong_thresh - INTEGER 1324 Fraction of packets marked with congestion over a round (RTT) to 1325 tag that round as congested. This is referred to as K in the PLB paper: 1326 https://doi.org/10.1145/3544216.3544226. 1327 1328 The 0-1 fraction range is mapped to 0-256 range to avoid floating 1329 point operations. For example, 128 means that if at least 50% of 1330 the packets in a round were marked as congested then the round 1331 will be tagged as congested. 1332 1333 Setting threshold to 0 means that PLB repaths every RTT regardless 1334 of congestion. This is not intended behavior for PLB and should be 1335 used only for experimentation purpose. 1336 1337 Possible Values: 0 - 256 1338 1339 Default: 128 1340 1341tcp_pingpong_thresh - INTEGER 1342 The number of estimated data replies sent for estimated incoming data 1343 requests that must happen before TCP considers that a connection is a 1344 "ping-pong" (request-response) connection for which delayed 1345 acknowledgments can provide benefits. 1346 1347 This threshold is 1 by default, but some applications may need a higher 1348 threshold for optimal performance. 1349 1350 Possible Values: 1 - 255 1351 1352 Default: 1 1353 1354tcp_rto_min_us - INTEGER 1355 Minimal TCP retransmission timeout (in microseconds). Note that the 1356 rto_min route option has the highest precedence for configuring this 1357 setting, followed by the TCP_BPF_RTO_MIN and TCP_RTO_MIN_US socket 1358 options, followed by this tcp_rto_min_us sysctl. 1359 1360 The recommended practice is to use a value less or equal to 200000 1361 microseconds. 1362 1363 Possible Values: 1 - INT_MAX 1364 1365 Default: 200000 1366 1367tcp_rto_max_ms - INTEGER 1368 Maximal TCP retransmission timeout (in ms). 1369 Note that TCP_RTO_MAX_MS socket option has higher precedence. 1370 1371 When changing tcp_rto_max_ms, it is important to understand 1372 that tcp_retries2 might need a change. 1373 1374 Possible Values: 1000 - 120,000 1375 1376 Default: 120,000 1377 1378UDP variables 1379============= 1380 1381udp_l3mdev_accept - BOOLEAN 1382 Enabling this option allows a "global" bound socket to work 1383 across L3 master domains (e.g., VRFs) with packets capable of 1384 being received regardless of the L3 domain in which they 1385 originated. Only valid when the kernel was compiled with 1386 CONFIG_NET_L3_MASTER_DEV. 1387 1388 Possible values: 1389 1390 - 0 (disabled) 1391 - 1 (enabled) 1392 1393 Default: 0 (disabled) 1394 1395udp_mem - vector of 3 INTEGERs: min, pressure, max 1396 Number of pages allowed for queueing by all UDP sockets. 1397 1398 min: Number of pages allowed for queueing by all UDP sockets. 1399 1400 pressure: This value was introduced to follow format of tcp_mem. 1401 1402 max: This value was introduced to follow format of tcp_mem. 1403 1404 Default is calculated at boot time from amount of available memory. 1405 1406udp_rmem_min - INTEGER 1407 Minimal size of receive buffer used by UDP sockets in moderation. 1408 Each UDP socket is able to use the size for receiving data, even if 1409 total pages of UDP sockets exceed udp_mem pressure. The unit is byte. 1410 1411 Default: 4K 1412 1413udp_wmem_min - INTEGER 1414 UDP does not have tx memory accounting and this tunable has no effect. 1415 1416udp_hash_entries - INTEGER 1417 Show the number of hash buckets for UDP sockets in the current 1418 networking namespace. 1419 1420 A negative value means the networking namespace does not own its 1421 hash buckets and shares the initial networking namespace's one. 1422 1423udp_child_hash_entries - INTEGER 1424 Control the number of hash buckets for UDP sockets in the child 1425 networking namespace, which must be set before clone() or unshare(). 1426 1427 If the value is not 0, the kernel uses a value rounded up to 2^n 1428 as the actual hash bucket size. 0 is a special value, meaning 1429 the child networking namespace will share the initial networking 1430 namespace's hash buckets. 1431 1432 Note that the child will use the global one in case the kernel 1433 fails to allocate enough memory. In addition, the global hash 1434 buckets are spread over available NUMA nodes, but the allocation 1435 of the child hash table depends on the current process's NUMA 1436 policy, which could result in performance differences. 1437 1438 Possible values: 0, 2^n (n: 7 (128) - 16 (64K)) 1439 1440 Default: 0 1441 1442 1443RAW variables 1444============= 1445 1446raw_l3mdev_accept - BOOLEAN 1447 Enabling this option allows a "global" bound socket to work 1448 across L3 master domains (e.g., VRFs) with packets capable of 1449 being received regardless of the L3 domain in which they 1450 originated. Only valid when the kernel was compiled with 1451 CONFIG_NET_L3_MASTER_DEV. 1452 1453 Possible values: 1454 1455 - 0 (disabled) 1456 - 1 (enabled) 1457 1458 Default: 1 (enabled) 1459 1460CIPSOv4 Variables 1461================= 1462 1463cipso_cache_enable - BOOLEAN 1464 If enabled, enable additions to and lookups from the CIPSO label mapping 1465 cache. If disabled, additions are ignored and lookups always result in a 1466 miss. However, regardless of the setting the cache is still 1467 invalidated when required when means you can safely toggle this on and 1468 off and the cache will always be "safe". 1469 1470 Possible values: 1471 1472 - 0 (disabled) 1473 - 1 (enabled) 1474 1475 Default: 1 (enabled) 1476 1477cipso_cache_bucket_size - INTEGER 1478 The CIPSO label cache consists of a fixed size hash table with each 1479 hash bucket containing a number of cache entries. This variable limits 1480 the number of entries in each hash bucket; the larger the value is, the 1481 more CIPSO label mappings that can be cached. When the number of 1482 entries in a given hash bucket reaches this limit adding new entries 1483 causes the oldest entry in the bucket to be removed to make room. 1484 1485 Default: 10 1486 1487cipso_rbm_optfmt - BOOLEAN 1488 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of 1489 the CIPSO draft specification (see Documentation/netlabel for details). 1490 This means that when set the CIPSO tag will be padded with empty 1491 categories in order to make the packet data 32-bit aligned. 1492 1493 Possible values: 1494 1495 - 0 (disabled) 1496 - 1 (enabled) 1497 1498 Default: 0 (disabled) 1499 1500cipso_rbm_strictvalid - BOOLEAN 1501 If enabled, do a very strict check of the CIPSO option when 1502 ip_options_compile() is called. If disabled, relax the checks done during 1503 ip_options_compile(). Either way is "safe" as errors are caught else 1504 where in the CIPSO processing code but setting this to 0 (False) should 1505 result in less work (i.e. it should be faster) but could cause problems 1506 with other implementations that require strict checking. 1507 1508 Possible values: 1509 1510 - 0 (disabled) 1511 - 1 (enabled) 1512 1513 Default: 0 (disabled) 1514 1515IP Variables 1516============ 1517 1518ip_local_port_range - 2 INTEGERS 1519 Defines the local port range that is used by TCP and UDP to 1520 choose the local port. The first number is the first, the 1521 second the last local port number. 1522 If possible, it is better these numbers have different parity 1523 (one even and one odd value). 1524 Must be greater than or equal to ip_unprivileged_port_start. 1525 The default values are 32768 and 60999 respectively. 1526 1527ip_local_reserved_ports - list of comma separated ranges 1528 Specify the ports which are reserved for known third-party 1529 applications. These ports will not be used by automatic port 1530 assignments (e.g. when calling connect() or bind() with port 1531 number 0). Explicit port allocation behavior is unchanged. 1532 1533 The format used for both input and output is a comma separated 1534 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and 1535 10). Writing to the file will clear all previously reserved 1536 ports and update the current list with the one given in the 1537 input. 1538 1539 Note that ip_local_port_range and ip_local_reserved_ports 1540 settings are independent and both are considered by the kernel 1541 when determining which ports are available for automatic port 1542 assignments. 1543 1544 You can reserve ports which are not in the current 1545 ip_local_port_range, e.g.:: 1546 1547 $ cat /proc/sys/net/ipv4/ip_local_port_range 1548 32000 60999 1549 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports 1550 8080,9148 1551 1552 although this is redundant. However such a setting is useful 1553 if later the port range is changed to a value that will 1554 include the reserved ports. Also keep in mind, that overlapping 1555 of these ranges may affect probability of selecting ephemeral 1556 ports which are right after block of reserved ports. 1557 1558 Default: Empty 1559 1560ip_unprivileged_port_start - INTEGER 1561 This is a per-namespace sysctl. It defines the first 1562 unprivileged port in the network namespace. Privileged ports 1563 require root or CAP_NET_BIND_SERVICE in order to bind to them. 1564 To disable all privileged ports, set this to 0. They must not 1565 overlap with the ip_local_port_range. 1566 1567 Default: 1024 1568 1569ip_nonlocal_bind - BOOLEAN 1570 If enabled, allows processes to bind() to non-local IP addresses, 1571 which can be quite useful - but may break some applications. 1572 1573 Possible values: 1574 1575 - 0 (disabled) 1576 - 1 (enabled) 1577 1578 Default: 0 (disabled) 1579 1580ip_autobind_reuse - BOOLEAN 1581 By default, bind() does not select the ports automatically even if 1582 the new socket and all sockets bound to the port have SO_REUSEADDR. 1583 ip_autobind_reuse allows bind() to reuse the port and this is useful 1584 when you use bind()+connect(), but may break some applications. 1585 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this 1586 option should only be set by experts. 1587 1588 Possible values: 1589 1590 - 0 (disabled) 1591 - 1 (enabled) 1592 1593 Default: 0 (disabled) 1594 1595ip_dynaddr - INTEGER 1596 If set non-zero, enables support for dynamic addresses. 1597 If set to a non-zero value larger than 1, a kernel log 1598 message will be printed when dynamic address rewriting 1599 occurs. 1600 1601 Default: 0 1602 1603ip_early_demux - BOOLEAN 1604 Optimize input packet processing down to one demux for 1605 certain kinds of local sockets. Currently we only do this 1606 for established TCP and connected UDP sockets. 1607 1608 It may add an additional cost for pure routing workloads that 1609 reduces overall throughput, in such case you should disable it. 1610 1611 Possible values: 1612 1613 - 0 (disabled) 1614 - 1 (enabled) 1615 1616 Default: 1 (enabled) 1617 1618ping_group_range - 2 INTEGERS 1619 Restrict ICMP_PROTO datagram sockets to users in the group range. 1620 The default is "1 0", meaning, that nobody (not even root) may 1621 create ping sockets. Setting it to "100 100" would grant permissions 1622 to the single group. "0 4294967294" would enable it for the world, "100 1623 4294967294" would enable it for the users, but not daemons. 1624 1625tcp_early_demux - BOOLEAN 1626 Enable early demux for established TCP sockets. 1627 1628 Possible values: 1629 1630 - 0 (disabled) 1631 - 1 (enabled) 1632 1633 Default: 1 (enabled) 1634 1635udp_early_demux - BOOLEAN 1636 Enable early demux for connected UDP sockets. Disable this if 1637 your system could experience more unconnected load. 1638 1639 Possible values: 1640 1641 - 0 (disabled) 1642 - 1 (enabled) 1643 1644 Default: 1 (enabled) 1645 1646icmp_echo_ignore_all - BOOLEAN 1647 If enabled, then the kernel will ignore all ICMP ECHO 1648 requests sent to it. 1649 1650 Possible values: 1651 1652 - 0 (disabled) 1653 - 1 (enabled) 1654 1655 Default: 0 (disabled) 1656 1657icmp_echo_enable_probe - BOOLEAN 1658 If enabled, then the kernel will respond to RFC 8335 PROBE 1659 requests sent to it. 1660 1661 Possible values: 1662 1663 - 0 (disabled) 1664 - 1 (enabled) 1665 1666 Default: 0 (disabled) 1667 1668icmp_echo_ignore_broadcasts - BOOLEAN 1669 If enabled, then the kernel will ignore all ICMP ECHO and 1670 TIMESTAMP requests sent to it via broadcast/multicast. 1671 1672 Possible values: 1673 1674 - 0 (disabled) 1675 - 1 (enabled) 1676 1677 Default: 1 (enabled) 1678 1679icmp_ratelimit - INTEGER 1680 Limit the maximal rates for sending ICMP packets whose type matches 1681 icmp_ratemask (see below) to specific targets. 1682 0 to disable any limiting, 1683 otherwise the minimal space between responses in milliseconds. 1684 Note that another sysctl, icmp_msgs_per_sec limits the number 1685 of ICMP packets sent on all targets. 1686 1687 Default: 1000 1688 1689icmp_msgs_per_sec - INTEGER 1690 Limit maximal number of ICMP packets sent per second from this host. 1691 Only messages whose type matches icmp_ratemask (see below) are 1692 controlled by this limit. For security reasons, the precise count 1693 of messages per second is randomized. 1694 1695 Default: 1000 1696 1697icmp_msgs_burst - INTEGER 1698 icmp_msgs_per_sec controls number of ICMP packets sent per second, 1699 while icmp_msgs_burst controls the burst size of these packets. 1700 For security reasons, the precise burst size is randomized. 1701 1702 Default: 50 1703 1704icmp_ratemask - INTEGER 1705 Mask made of ICMP types for which rates are being limited. 1706 1707 Significant bits: IHGFEDCBA9876543210 1708 1709 Default mask: 0000001100000011000 (6168) 1710 1711 Bit definitions (see include/linux/icmp.h): 1712 1713 = ========================= 1714 0 Echo Reply 1715 3 Destination Unreachable [1]_ 1716 4 Source Quench [1]_ 1717 5 Redirect 1718 8 Echo Request 1719 B Time Exceeded [1]_ 1720 C Parameter Problem [1]_ 1721 D Timestamp Request 1722 E Timestamp Reply 1723 F Info Request 1724 G Info Reply 1725 H Address Mask Request 1726 I Address Mask Reply 1727 = ========================= 1728 1729 .. [1] These are rate limited by default (see default mask above) 1730 1731icmp_ignore_bogus_error_responses - BOOLEAN 1732 Some routers violate RFC1122 by sending bogus responses to broadcast 1733 frames. Such violations are normally logged via a kernel warning. 1734 If enabled, the kernel will not give such warnings, which 1735 will avoid log file clutter. 1736 1737 Possible values: 1738 1739 - 0 (disabled) 1740 - 1 (enabled) 1741 1742 Default: 1 (enabled) 1743 1744icmp_errors_use_inbound_ifaddr - BOOLEAN 1745 1746 If disabled, icmp error messages are sent with the primary address of 1747 the exiting interface. 1748 1749 If enabled, the message will be sent with the primary address of 1750 the interface that received the packet that caused the icmp error. 1751 This is the behaviour many network administrators will expect from 1752 a router. And it can make debugging complicated network layouts 1753 much easier. 1754 1755 Note that if no primary address exists for the interface selected, 1756 then the primary address of the first non-loopback interface that 1757 has one will be used regardless of this setting. 1758 1759 Possible values: 1760 1761 - 0 (disabled) 1762 - 1 (enabled) 1763 1764 Default: 0 (disabled) 1765 1766igmp_max_memberships - INTEGER 1767 Change the maximum number of multicast groups we can subscribe to. 1768 Default: 20 1769 1770 Theoretical maximum value is bounded by having to send a membership 1771 report in a single datagram (i.e. the report can't span multiple 1772 datagrams, or risk confusing the switch and leaving groups you don't 1773 intend to). 1774 1775 The number of supported groups 'M' is bounded by the number of group 1776 report entries you can fit into a single datagram of 65535 bytes. 1777 1778 M = 65536-sizeof (ip header)/(sizeof(Group record)) 1779 1780 Group records are variable length, with a minimum of 12 bytes. 1781 So net.ipv4.igmp_max_memberships should not be set higher than: 1782 1783 (65536-24) / 12 = 5459 1784 1785 The value 5459 assumes no IP header options, so in practice 1786 this number may be lower. 1787 1788igmp_max_msf - INTEGER 1789 Maximum number of addresses allowed in the source filter list for a 1790 multicast group. 1791 1792 Default: 10 1793 1794igmp_qrv - INTEGER 1795 Controls the IGMP query robustness variable (see RFC2236 8.1). 1796 1797 Default: 2 (as specified by RFC2236 8.1) 1798 1799 Minimum: 1 (as specified by RFC6636 4.5) 1800 1801force_igmp_version - INTEGER 1802 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback 1803 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier 1804 Present timer expires. 1805 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if 1806 receive IGMPv2/v3 query. 1807 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive 1808 IGMPv1 query message. Will reply report if receive IGMPv3 query. 1809 - 3 - Enforce to use IGMP version 3. The same react with default 0. 1810 1811 .. note:: 1812 1813 this is not the same with force_mld_version because IGMPv3 RFC3376 1814 Security Considerations does not have clear description that we could 1815 ignore other version messages completely as MLDv2 RFC3810. So make 1816 this value as default 0 is recommended. 1817 1818``conf/interface/*`` 1819 changes special settings per interface (where 1820 interface" is the name of your network interface) 1821 1822``conf/all/*`` 1823 is special, changes the settings for all interfaces 1824 1825log_martians - BOOLEAN 1826 Log packets with impossible addresses to kernel log. 1827 log_martians for the interface will be enabled if at least one of 1828 conf/{all,interface}/log_martians is set to TRUE, 1829 it will be disabled otherwise 1830 1831accept_redirects - BOOLEAN 1832 Accept ICMP redirect messages. 1833 accept_redirects for the interface will be enabled if: 1834 1835 - both conf/{all,interface}/accept_redirects are TRUE in the case 1836 forwarding for the interface is enabled 1837 1838 or 1839 1840 - at least one of conf/{all,interface}/accept_redirects is TRUE in the 1841 case forwarding for the interface is disabled 1842 1843 accept_redirects for the interface will be disabled otherwise 1844 1845 default: 1846 1847 - TRUE (host) 1848 - FALSE (router) 1849 1850forwarding - BOOLEAN 1851 Enable IP forwarding on this interface. This controls whether packets 1852 received _on_ this interface can be forwarded. 1853 1854mc_forwarding - BOOLEAN 1855 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE 1856 and a multicast routing daemon is required. 1857 conf/all/mc_forwarding must also be set to TRUE to enable multicast 1858 routing for the interface 1859 1860medium_id - INTEGER 1861 Integer value used to differentiate the devices by the medium they 1862 are attached to. Two devices can have different id values when 1863 the broadcast packets are received only on one of them. 1864 The default value 0 means that the device is the only interface 1865 to its medium, value of -1 means that medium is not known. 1866 1867 Currently, it is used to change the proxy_arp behavior: 1868 the proxy_arp feature is enabled for packets forwarded between 1869 two devices attached to different media. 1870 1871proxy_arp - BOOLEAN 1872 Do proxy arp. 1873 1874 proxy_arp for the interface will be enabled if at least one of 1875 conf/{all,interface}/proxy_arp is set to TRUE, 1876 it will be disabled otherwise 1877 1878proxy_arp_pvlan - BOOLEAN 1879 Private VLAN proxy arp. 1880 1881 Basically allow proxy arp replies back to the same interface 1882 (from which the ARP request/solicitation was received). 1883 1884 This is done to support (ethernet) switch features, like RFC 1885 3069, where the individual ports are NOT allowed to 1886 communicate with each other, but they are allowed to talk to 1887 the upstream router. As described in RFC 3069, it is possible 1888 to allow these hosts to communicate through the upstream 1889 router by proxy_arp'ing. Don't need to be used together with 1890 proxy_arp. 1891 1892 This technology is known by different names: 1893 1894 - In RFC 3069 it is called VLAN Aggregation. 1895 - Cisco and Allied Telesyn call it Private VLAN. 1896 - Hewlett-Packard call it Source-Port filtering or port-isolation. 1897 - Ericsson call it MAC-Forced Forwarding (RFC Draft). 1898 1899proxy_delay - INTEGER 1900 Delay proxy response. 1901 1902 Delay response to a neighbor solicitation when proxy_arp 1903 or proxy_ndp is enabled. A random value between [0, proxy_delay) 1904 will be chosen, setting to zero means reply with no delay. 1905 Value in jiffies. Defaults to 80. 1906 1907shared_media - BOOLEAN 1908 Send(router) or accept(host) RFC1620 shared media redirects. 1909 Overrides secure_redirects. 1910 1911 shared_media for the interface will be enabled if at least one of 1912 conf/{all,interface}/shared_media is set to TRUE, 1913 it will be disabled otherwise 1914 1915 default TRUE 1916 1917secure_redirects - BOOLEAN 1918 Accept ICMP redirect messages only to gateways listed in the 1919 interface's current gateway list. Even if disabled, RFC1122 redirect 1920 rules still apply. 1921 1922 Overridden by shared_media. 1923 1924 secure_redirects for the interface will be enabled if at least one of 1925 conf/{all,interface}/secure_redirects is set to TRUE, 1926 it will be disabled otherwise 1927 1928 default TRUE 1929 1930send_redirects - BOOLEAN 1931 Send redirects, if router. 1932 1933 send_redirects for the interface will be enabled if at least one of 1934 conf/{all,interface}/send_redirects is set to TRUE, 1935 it will be disabled otherwise 1936 1937 Default: TRUE 1938 1939bootp_relay - BOOLEAN 1940 Accept packets with source address 0.b.c.d destined 1941 not to this host as local ones. It is supposed, that 1942 BOOTP relay daemon will catch and forward such packets. 1943 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay 1944 for the interface 1945 1946 default FALSE 1947 1948 Not Implemented Yet. 1949 1950accept_source_route - BOOLEAN 1951 Accept packets with SRR option. 1952 conf/all/accept_source_route must also be set to TRUE to accept packets 1953 with SRR option on the interface 1954 1955 default 1956 1957 - TRUE (router) 1958 - FALSE (host) 1959 1960accept_local - BOOLEAN 1961 Accept packets with local source addresses. In combination with 1962 suitable routing, this can be used to direct packets between two 1963 local interfaces over the wire and have them accepted properly. 1964 default FALSE 1965 1966route_localnet - BOOLEAN 1967 Do not consider loopback addresses as martian source or destination 1968 while routing. This enables the use of 127/8 for local routing purposes. 1969 1970 default FALSE 1971 1972rp_filter - INTEGER 1973 - 0 - No source validation. 1974 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path 1975 Each incoming packet is tested against the FIB and if the interface 1976 is not the best reverse path the packet check will fail. 1977 By default failed packets are discarded. 1978 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path 1979 Each incoming packet's source address is also tested against the FIB 1980 and if the source address is not reachable via any interface 1981 the packet check will fail. 1982 1983 Current recommended practice in RFC3704 is to enable strict mode 1984 to prevent IP spoofing from DDos attacks. If using asymmetric routing 1985 or other complicated routing, then loose mode is recommended. 1986 1987 The max value from conf/{all,interface}/rp_filter is used 1988 when doing source validation on the {interface}. 1989 1990 Default value is 0. Note that some distributions enable it 1991 in startup scripts. 1992 1993src_valid_mark - BOOLEAN 1994 - 0 - The fwmark of the packet is not included in reverse path 1995 route lookup. This allows for asymmetric routing configurations 1996 utilizing the fwmark in only one direction, e.g., transparent 1997 proxying. 1998 1999 - 1 - The fwmark of the packet is included in reverse path route 2000 lookup. This permits rp_filter to function when the fwmark is 2001 used for routing traffic in both directions. 2002 2003 This setting also affects the utilization of fmwark when 2004 performing source address selection for ICMP replies, or 2005 determining addresses stored for the IPOPT_TS_TSANDADDR and 2006 IPOPT_RR IP options. 2007 2008 The max value from conf/{all,interface}/src_valid_mark is used. 2009 2010 Default value is 0. 2011 2012arp_filter - BOOLEAN 2013 - 1 - Allows you to have multiple network interfaces on the same 2014 subnet, and have the ARPs for each interface be answered 2015 based on whether or not the kernel would route a packet from 2016 the ARP'd IP out that interface (therefore you must use source 2017 based routing for this to work). In other words it allows control 2018 of which cards (usually 1) will respond to an arp request. 2019 2020 - 0 - (default) The kernel can respond to arp requests with addresses 2021 from other interfaces. This may seem wrong but it usually makes 2022 sense, because it increases the chance of successful communication. 2023 IP addresses are owned by the complete host on Linux, not by 2024 particular interfaces. Only for more complex setups like load- 2025 balancing, does this behaviour cause problems. 2026 2027 arp_filter for the interface will be enabled if at least one of 2028 conf/{all,interface}/arp_filter is set to TRUE, 2029 it will be disabled otherwise 2030 2031arp_announce - INTEGER 2032 Define different restriction levels for announcing the local 2033 source IP address from IP packets in ARP requests sent on 2034 interface: 2035 2036 - 0 - (default) Use any local address, configured on any interface 2037 - 1 - Try to avoid local addresses that are not in the target's 2038 subnet for this interface. This mode is useful when target 2039 hosts reachable via this interface require the source IP 2040 address in ARP requests to be part of their logical network 2041 configured on the receiving interface. When we generate the 2042 request we will check all our subnets that include the 2043 target IP and will preserve the source address if it is from 2044 such subnet. If there is no such subnet we select source 2045 address according to the rules for level 2. 2046 - 2 - Always use the best local address for this target. 2047 In this mode we ignore the source address in the IP packet 2048 and try to select local address that we prefer for talks with 2049 the target host. Such local address is selected by looking 2050 for primary IP addresses on all our subnets on the outgoing 2051 interface that include the target IP address. If no suitable 2052 local address is found we select the first local address 2053 we have on the outgoing interface or on all other interfaces, 2054 with the hope we will receive reply for our request and 2055 even sometimes no matter the source IP address we announce. 2056 2057 The max value from conf/{all,interface}/arp_announce is used. 2058 2059 Increasing the restriction level gives more chance for 2060 receiving answer from the resolved target while decreasing 2061 the level announces more valid sender's information. 2062 2063arp_ignore - INTEGER 2064 Define different modes for sending replies in response to 2065 received ARP requests that resolve local target IP addresses: 2066 2067 - 0 - (default): reply for any local target IP address, configured 2068 on any interface 2069 - 1 - reply only if the target IP address is local address 2070 configured on the incoming interface 2071 - 2 - reply only if the target IP address is local address 2072 configured on the incoming interface and both with the 2073 sender's IP address are part from same subnet on this interface 2074 - 3 - do not reply for local addresses configured with scope host, 2075 only resolutions for global and link addresses are replied 2076 - 4-7 - reserved 2077 - 8 - do not reply for all local addresses 2078 2079 The max value from conf/{all,interface}/arp_ignore is used 2080 when ARP request is received on the {interface} 2081 2082arp_notify - BOOLEAN 2083 Define mode for notification of address and device changes. 2084 2085 == ========================================================== 2086 0 (default): do nothing 2087 1 Generate gratuitous arp requests when device is brought up 2088 or hardware address changes. 2089 == ========================================================== 2090 2091arp_accept - INTEGER 2092 Define behavior for accepting gratuitous ARP (garp) frames from devices 2093 that are not already present in the ARP table: 2094 2095 - 0 - don't create new entries in the ARP table 2096 - 1 - create new entries in the ARP table 2097 - 2 - create new entries only if the source IP address is in the same 2098 subnet as an address configured on the interface that received the 2099 garp message. 2100 2101 Both replies and requests type gratuitous arp will trigger the 2102 ARP table to be updated, if this setting is on. 2103 2104 If the ARP table already contains the IP address of the 2105 gratuitous arp frame, the arp table will be updated regardless 2106 if this setting is on or off. 2107 2108arp_evict_nocarrier - BOOLEAN 2109 Clears the ARP cache on NOCARRIER events. This option is important for 2110 wireless devices where the ARP cache should not be cleared when roaming 2111 between access points on the same network. In most cases this should 2112 remain as the default (1). 2113 2114 Possible values: 2115 2116 - 0 (disabled) - Do not clear ARP cache on NOCARRIER events 2117 - 1 (enabled) - Clear the ARP cache on NOCARRIER events 2118 2119 Default: 1 (enabled) 2120 2121mcast_solicit - INTEGER 2122 The maximum number of multicast probes in INCOMPLETE state, 2123 when the associated hardware address is unknown. Defaults 2124 to 3. 2125 2126ucast_solicit - INTEGER 2127 The maximum number of unicast probes in PROBE state, when 2128 the hardware address is being reconfirmed. Defaults to 3. 2129 2130app_solicit - INTEGER 2131 The maximum number of probes to send to the user space ARP daemon 2132 via netlink before dropping back to multicast probes (see 2133 mcast_resolicit). Defaults to 0. 2134 2135mcast_resolicit - INTEGER 2136 The maximum number of multicast probes after unicast and 2137 app probes in PROBE state. Defaults to 0. 2138 2139disable_policy - BOOLEAN 2140 Disable IPSEC policy (SPD) for this interface 2141 2142 Possible values: 2143 2144 - 0 (disabled) 2145 - 1 (enabled) 2146 2147 Default: 0 (disabled) 2148 2149disable_xfrm - BOOLEAN 2150 Disable IPSEC encryption on this interface, whatever the policy 2151 2152 Possible values: 2153 2154 - 0 (disabled) 2155 - 1 (enabled) 2156 2157 Default: 0 (disabled) 2158 2159igmpv2_unsolicited_report_interval - INTEGER 2160 The interval in milliseconds in which the next unsolicited 2161 IGMPv1 or IGMPv2 report retransmit will take place. 2162 2163 Default: 10000 (10 seconds) 2164 2165igmpv3_unsolicited_report_interval - INTEGER 2166 The interval in milliseconds in which the next unsolicited 2167 IGMPv3 report retransmit will take place. 2168 2169 Default: 1000 (1 seconds) 2170 2171ignore_routes_with_linkdown - BOOLEAN 2172 Ignore routes whose link is down when performing a FIB lookup. 2173 2174 Possible values: 2175 2176 - 0 (disabled) 2177 - 1 (enabled) 2178 2179 Default: 0 (disabled) 2180 2181promote_secondaries - BOOLEAN 2182 When a primary IP address is removed from this interface 2183 promote a corresponding secondary IP address instead of 2184 removing all the corresponding secondary IP addresses. 2185 2186 Possible values: 2187 2188 - 0 (disabled) 2189 - 1 (enabled) 2190 2191 Default: 0 (disabled) 2192 2193drop_unicast_in_l2_multicast - BOOLEAN 2194 Drop any unicast IP packets that are received in link-layer 2195 multicast (or broadcast) frames. 2196 2197 This behavior (for multicast) is actually a SHOULD in RFC 2198 1122, but is disabled by default for compatibility reasons. 2199 2200 Possible values: 2201 2202 - 0 (disabled) 2203 - 1 (enabled) 2204 2205 Default: 0 (disabled) 2206 2207drop_gratuitous_arp - BOOLEAN 2208 Drop all gratuitous ARP frames, for example if there's a known 2209 good ARP proxy on the network and such frames need not be used 2210 (or in the case of 802.11, must not be used to prevent attacks.) 2211 2212 Possible values: 2213 2214 - 0 (disabled) 2215 - 1 (enabled) 2216 2217 Default: 0 (disabled) 2218 2219 2220tag - INTEGER 2221 Allows you to write a number, which can be used as required. 2222 2223 Default value is 0. 2224 2225xfrm4_gc_thresh - INTEGER 2226 (Obsolete since linux-4.14) 2227 The threshold at which we will start garbage collecting for IPv4 2228 destination cache entries. At twice this value the system will 2229 refuse new allocations. 2230 2231igmp_link_local_mcast_reports - BOOLEAN 2232 Enable IGMP reports for link local multicast groups in the 2233 224.0.0.X range. 2234 2235 Default TRUE 2236 2237Alexey Kuznetsov. 2238kuznet@ms2.inr.ac.ru 2239 2240Updated by: 2241 2242- Andi Kleen 2243 ak@muc.de 2244- Nicolas Delon 2245 delon.nicolas@wanadoo.fr 2246 2247 2248 2249 2250/proc/sys/net/ipv6/* Variables 2251============================== 2252 2253IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also 2254apply to IPv6 [XXX?]. 2255 2256bindv6only - BOOLEAN 2257 Default value for IPV6_V6ONLY socket option, 2258 which restricts use of the IPv6 socket to IPv6 communication 2259 only. 2260 2261 Possible values: 2262 2263 - 0 (disabled) - enable IPv4-mapped address feature 2264 - 1 (enabled) - disable IPv4-mapped address feature 2265 2266 Default: 0 (disabled) 2267 2268flowlabel_consistency - BOOLEAN 2269 Protect the consistency (and unicity) of flow label. 2270 You have to disable it to use IPV6_FL_F_REFLECT flag on the 2271 flow label manager. 2272 2273 Possible values: 2274 2275 - 0 (disabled) 2276 - 1 (enabled) 2277 2278 Default: 1 (enabled) 2279 2280auto_flowlabels - INTEGER 2281 Automatically generate flow labels based on a flow hash of the 2282 packet. This allows intermediate devices, such as routers, to 2283 identify packet flows for mechanisms like Equal Cost Multipath 2284 Routing (see RFC 6438). 2285 2286 = =========================================================== 2287 0 automatic flow labels are completely disabled 2288 1 automatic flow labels are enabled by default, they can be 2289 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL 2290 socket option 2291 2 automatic flow labels are allowed, they may be enabled on a 2292 per socket basis using the IPV6_AUTOFLOWLABEL socket option 2293 3 automatic flow labels are enabled and enforced, they cannot 2294 be disabled by the socket option 2295 = =========================================================== 2296 2297 Default: 1 2298 2299flowlabel_state_ranges - BOOLEAN 2300 Split the flow label number space into two ranges. 0-0x7FFFF is 2301 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF 2302 is reserved for stateless flow labels as described in RFC6437. 2303 2304 Possible values: 2305 2306 - 0 (disabled) 2307 - 1 (enabled) 2308 2309 Default: 1 (enabled) 2310 2311 2312flowlabel_reflect - INTEGER 2313 Control flow label reflection. Needed for Path MTU 2314 Discovery to work with Equal Cost Multipath Routing in anycast 2315 environments. See RFC 7690 and: 2316 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01 2317 2318 This is a bitmask. 2319 2320 - 1: enabled for established flows 2321 2322 Note that this prevents automatic flowlabel changes, as done 2323 in "tcp: change IPv6 flow-label upon receiving spurious retransmission" 2324 and "tcp: Change txhash on every SYN and RTO retransmit" 2325 2326 - 2: enabled for TCP RESET packets (no active listener) 2327 If set, a RST packet sent in response to a SYN packet on a closed 2328 port will reflect the incoming flow label. 2329 2330 - 4: enabled for ICMPv6 echo reply messages. 2331 2332 Default: 0 2333 2334fib_multipath_hash_policy - INTEGER 2335 Controls which hash policy to use for multipath routes. 2336 2337 Default: 0 (Layer 3) 2338 2339 Possible values: 2340 2341 - 0 - Layer 3 (source and destination addresses plus flow label) 2342 - 1 - Layer 4 (standard 5-tuple) 2343 - 2 - Layer 3 or inner Layer 3 if present 2344 - 3 - Custom multipath hash. Fields used for multipath hash calculation 2345 are determined by fib_multipath_hash_fields sysctl 2346 2347fib_multipath_hash_fields - UNSIGNED INTEGER 2348 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the 2349 fields used for multipath hash calculation are determined by this 2350 sysctl. 2351 2352 This value is a bitmask which enables various fields for multipath hash 2353 calculation. 2354 2355 Possible fields are: 2356 2357 ====== ============================ 2358 0x0001 Source IP address 2359 0x0002 Destination IP address 2360 0x0004 IP protocol 2361 0x0008 Flow Label 2362 0x0010 Source port 2363 0x0020 Destination port 2364 0x0040 Inner source IP address 2365 0x0080 Inner destination IP address 2366 0x0100 Inner IP protocol 2367 0x0200 Inner Flow Label 2368 0x0400 Inner source port 2369 0x0800 Inner destination port 2370 ====== ============================ 2371 2372 Default: 0x0007 (source IP, destination IP and IP protocol) 2373 2374anycast_src_echo_reply - BOOLEAN 2375 Controls the use of anycast addresses as source addresses for ICMPv6 2376 echo reply 2377 2378 Possible values: 2379 2380 - 0 (disabled) 2381 - 1 (enabled) 2382 2383 Default: 0 (disabled) 2384 2385 2386idgen_delay - INTEGER 2387 Controls the delay in seconds after which time to retry 2388 privacy stable address generation if a DAD conflict is 2389 detected. 2390 2391 Default: 1 (as specified in RFC7217) 2392 2393idgen_retries - INTEGER 2394 Controls the number of retries to generate a stable privacy 2395 address if a DAD conflict is detected. 2396 2397 Default: 3 (as specified in RFC7217) 2398 2399mld_qrv - INTEGER 2400 Controls the MLD query robustness variable (see RFC3810 9.1). 2401 2402 Default: 2 (as specified by RFC3810 9.1) 2403 2404 Minimum: 1 (as specified by RFC6636 4.5) 2405 2406max_dst_opts_number - INTEGER 2407 Maximum number of non-padding TLVs allowed in a Destination 2408 options extension header. If this value is less than zero 2409 then unknown options are disallowed and the number of known 2410 TLVs allowed is the absolute value of this number. 2411 2412 Default: 8 2413 2414max_hbh_opts_number - INTEGER 2415 Maximum number of non-padding TLVs allowed in a Hop-by-Hop 2416 options extension header. If this value is less than zero 2417 then unknown options are disallowed and the number of known 2418 TLVs allowed is the absolute value of this number. 2419 2420 Default: 8 2421 2422max_dst_opts_length - INTEGER 2423 Maximum length allowed for a Destination options extension 2424 header. 2425 2426 Default: INT_MAX (unlimited) 2427 2428max_hbh_length - INTEGER 2429 Maximum length allowed for a Hop-by-Hop options extension 2430 header. 2431 2432 Default: INT_MAX (unlimited) 2433 2434skip_notify_on_dev_down - BOOLEAN 2435 Controls whether an RTM_DELROUTE message is generated for routes 2436 removed when a device is taken down or deleted. IPv4 does not 2437 generate this message; IPv6 does by default. Setting this sysctl 2438 to true skips the message, making IPv4 and IPv6 on par in relying 2439 on userspace caches to track link events and evict routes. 2440 2441 Possible values: 2442 2443 - 0 (disabled) - generate the message 2444 - 1 (enabled) - skip generating the message 2445 2446 Default: 0 (disabled) 2447 2448nexthop_compat_mode - BOOLEAN 2449 New nexthop API provides a means for managing nexthops independent of 2450 prefixes. Backwards compatibility with old route format is enabled by 2451 default which means route dumps and notifications contain the new 2452 nexthop attribute but also the full, expanded nexthop definition. 2453 Further, updates or deletes of a nexthop configuration generate route 2454 notifications for each fib entry using the nexthop. Once a system 2455 understands the new API, this sysctl can be disabled to achieve full 2456 performance benefits of the new API by disabling the nexthop expansion 2457 and extraneous notifications. 2458 2459 Note that as a backward-compatible mode, dumping of modern features 2460 might be incomplete or wrong. For example, resilient groups will not be 2461 shown as such, but rather as just a list of next hops. Also weights that 2462 do not fit into 8 bits will show incorrectly. 2463 2464 Default: true (backward compat mode) 2465 2466fib_notify_on_flag_change - INTEGER 2467 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/ 2468 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed. 2469 2470 After installing a route to the kernel, user space receives an 2471 acknowledgment, which means the route was installed in the kernel, 2472 but not necessarily in hardware. 2473 It is also possible for a route already installed in hardware to change 2474 its action and therefore its flags. For example, a host route that is 2475 trapping packets can be "promoted" to perform decapsulation following 2476 the installation of an IPinIP/VXLAN tunnel. 2477 The notifications will indicate to user-space the state of the route. 2478 2479 Default: 0 (Do not emit notifications.) 2480 2481 Possible values: 2482 2483 - 0 - Do not emit notifications. 2484 - 1 - Emit notifications. 2485 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change. 2486 2487ioam6_id - INTEGER 2488 Define the IOAM id of this node. Uses only 24 bits out of 32 in total. 2489 2490 Possible value range: 2491 2492 - Min: 0 2493 - Max: 0xFFFFFF 2494 2495 Default: 0xFFFFFF 2496 2497ioam6_id_wide - LONG INTEGER 2498 Define the wide IOAM id of this node. Uses only 56 bits out of 64 in 2499 total. Can be different from ioam6_id. 2500 2501 Possible value range: 2502 2503 - Min: 0 2504 - Max: 0xFFFFFFFFFFFFFF 2505 2506 Default: 0xFFFFFFFFFFFFFF 2507 2508IPv6 Fragmentation: 2509 2510ip6frag_high_thresh - INTEGER 2511 Maximum memory used to reassemble IPv6 fragments. When 2512 ip6frag_high_thresh bytes of memory is allocated for this purpose, 2513 the fragment handler will toss packets until ip6frag_low_thresh 2514 is reached. 2515 2516ip6frag_low_thresh - INTEGER 2517 See ip6frag_high_thresh 2518 2519ip6frag_time - INTEGER 2520 Time in seconds to keep an IPv6 fragment in memory. 2521 2522``conf/default/*``: 2523 Change the interface-specific default settings. 2524 2525 These settings would be used during creating new interfaces. 2526 2527 2528``conf/all/*``: 2529 Change all the interface-specific settings. 2530 2531 [XXX: Other special features than forwarding?] 2532 2533conf/all/disable_ipv6 - BOOLEAN 2534 Changing this value is same as changing ``conf/default/disable_ipv6`` 2535 setting and also all per-interface ``disable_ipv6`` settings to the same 2536 value. 2537 2538 Reading this value does not have any particular meaning. It does not say 2539 whether IPv6 support is enabled or disabled. Returned value can be 1 2540 also in the case when some interface has ``disable_ipv6`` set to 0 and 2541 has configured IPv6 addresses. 2542 2543conf/all/forwarding - BOOLEAN 2544 Enable global IPv6 forwarding between all interfaces. 2545 2546 IPv4 and IPv6 work differently here; the ``force_forwarding`` flag must 2547 be used to control which interfaces may forward packets. 2548 2549 This also sets all interfaces' Host/Router setting 2550 'forwarding' to the specified value. See below for details. 2551 2552 This referred to as global forwarding. 2553 2554proxy_ndp - BOOLEAN 2555 Do proxy ndp. 2556 2557 Possible values: 2558 2559 - 0 (disabled) 2560 - 1 (enabled) 2561 2562 Default: 0 (disabled) 2563 2564force_forwarding - BOOLEAN 2565 Enable forwarding on this interface only -- regardless of the setting on 2566 ``conf/all/forwarding``. When setting ``conf.all.forwarding`` to 0, 2567 the ``force_forwarding`` flag will be reset on all interfaces. 2568 2569fwmark_reflect - BOOLEAN 2570 Controls the fwmark of kernel-generated IPv6 reply packets that are not 2571 associated with a socket for example, TCP RSTs or ICMPv6 echo replies). 2572 If disabled, these packets have a fwmark of zero. If enabled, they have the 2573 fwmark of the packet they are replying to. 2574 2575 Possible values: 2576 2577 - 0 (disabled) 2578 - 1 (enabled) 2579 2580 Default: 0 (disabled) 2581 2582``conf/interface/*``: 2583 Change special settings per interface. 2584 2585 The functional behaviour for certain settings is different 2586 depending on whether local forwarding is enabled or not. 2587 2588accept_ra - INTEGER 2589 Accept Router Advertisements; autoconfigure using them. 2590 2591 It also determines whether or not to transmit Router 2592 Solicitations. If and only if the functional setting is to 2593 accept Router Advertisements, Router Solicitations will be 2594 transmitted. 2595 2596 Possible values are: 2597 2598 == =========================================================== 2599 0 Do not accept Router Advertisements. 2600 1 Accept Router Advertisements if forwarding is disabled. 2601 2 Overrule forwarding behaviour. Accept Router Advertisements 2602 even if forwarding is enabled. 2603 == =========================================================== 2604 2605 Functional default: 2606 2607 - enabled if local forwarding is disabled. 2608 - disabled if local forwarding is enabled. 2609 2610accept_ra_defrtr - BOOLEAN 2611 Learn default router in Router Advertisement. 2612 2613 Functional default: 2614 2615 - enabled if accept_ra is enabled. 2616 - disabled if accept_ra is disabled. 2617 2618ra_defrtr_metric - UNSIGNED INTEGER 2619 Route metric for default route learned in Router Advertisement. This value 2620 will be assigned as metric for the default route learned via IPv6 Router 2621 Advertisement. Takes affect only if accept_ra_defrtr is enabled. 2622 2623 Possible values: 2624 1 to 0xFFFFFFFF 2625 2626 Default: IP6_RT_PRIO_USER i.e. 1024. 2627 2628accept_ra_from_local - BOOLEAN 2629 Accept RA with source-address that is found on local machine 2630 if the RA is otherwise proper and able to be accepted. 2631 2632 Default is to NOT accept these as it may be an un-intended 2633 network loop. 2634 2635 Functional default: 2636 2637 - enabled if accept_ra_from_local is enabled 2638 on a specific interface. 2639 - disabled if accept_ra_from_local is disabled 2640 on a specific interface. 2641 2642accept_ra_min_hop_limit - INTEGER 2643 Minimum hop limit Information in Router Advertisement. 2644 2645 Hop limit Information in Router Advertisement less than this 2646 variable shall be ignored. 2647 2648 Default: 1 2649 2650accept_ra_min_lft - INTEGER 2651 Minimum acceptable lifetime value in Router Advertisement. 2652 2653 RA sections with a lifetime less than this value shall be 2654 ignored. Zero lifetimes stay unaffected. 2655 2656 Default: 0 2657 2658accept_ra_pinfo - BOOLEAN 2659 Learn Prefix Information in Router Advertisement. 2660 2661 Functional default: 2662 2663 - enabled if accept_ra is enabled. 2664 - disabled if accept_ra is disabled. 2665 2666ra_honor_pio_life - BOOLEAN 2667 Whether to use RFC4862 Section 5.5.3e to determine the valid 2668 lifetime of an address matching a prefix sent in a Router 2669 Advertisement Prefix Information Option. 2670 2671 Possible values: 2672 2673 - 0 (disabled) - RFC4862 section 5.5.3e is used to determine 2674 the valid lifetime of the address. 2675 - 1 (enabled) - the PIO valid lifetime will always be honored. 2676 2677 Default: 0 (disabled) 2678 2679ra_honor_pio_pflag - BOOLEAN 2680 The Prefix Information Option P-flag indicates the network can 2681 allocate a unique IPv6 prefix per client using DHCPv6-PD. 2682 This sysctl can be enabled when a userspace DHCPv6-PD client 2683 is running to cause the P-flag to take effect: i.e. the 2684 P-flag suppresses any effects of the A-flag within the same 2685 PIO. For a given PIO, P=1 and A=1 is treated as A=0. 2686 2687 Possible values: 2688 2689 - 0 (disabled) - the P-flag is ignored. 2690 - 1 (enabled) - the P-flag will disable SLAAC autoconfiguration 2691 for the given Prefix Information Option. 2692 2693 Default: 0 (disabled) 2694 2695accept_ra_rt_info_min_plen - INTEGER 2696 Minimum prefix length of Route Information in RA. 2697 2698 Route Information w/ prefix smaller than this variable shall 2699 be ignored. 2700 2701 Functional default: 2702 2703 * 0 if accept_ra_rtr_pref is enabled. 2704 * -1 if accept_ra_rtr_pref is disabled. 2705 2706accept_ra_rt_info_max_plen - INTEGER 2707 Maximum prefix length of Route Information in RA. 2708 2709 Route Information w/ prefix larger than this variable shall 2710 be ignored. 2711 2712 Functional default: 2713 2714 * 0 if accept_ra_rtr_pref is enabled. 2715 * -1 if accept_ra_rtr_pref is disabled. 2716 2717accept_ra_rtr_pref - BOOLEAN 2718 Accept Router Preference in RA. 2719 2720 Functional default: 2721 2722 - enabled if accept_ra is enabled. 2723 - disabled if accept_ra is disabled. 2724 2725accept_ra_mtu - BOOLEAN 2726 Apply the MTU value specified in RA option 5 (RFC4861). If 2727 disabled, the MTU specified in the RA will be ignored. 2728 2729 Functional default: 2730 2731 - enabled if accept_ra is enabled. 2732 - disabled if accept_ra is disabled. 2733 2734accept_redirects - BOOLEAN 2735 Accept Redirects. 2736 2737 Functional default: 2738 2739 - enabled if local forwarding is disabled. 2740 - disabled if local forwarding is enabled. 2741 2742accept_source_route - INTEGER 2743 Accept source routing (routing extension header). 2744 2745 - >= 0: Accept only routing header type 2. 2746 - < 0: Do not accept routing header. 2747 2748 Default: 0 2749 2750autoconf - BOOLEAN 2751 Autoconfigure addresses using Prefix Information in Router 2752 Advertisements. 2753 2754 Functional default: 2755 2756 - enabled if accept_ra_pinfo is enabled. 2757 - disabled if accept_ra_pinfo is disabled. 2758 2759dad_transmits - INTEGER 2760 The amount of Duplicate Address Detection probes to send. 2761 2762 Default: 1 2763 2764forwarding - INTEGER 2765 Configure interface-specific Host/Router behaviour. 2766 2767 .. note:: 2768 2769 It is recommended to have the same setting on all 2770 interfaces; mixed router/host scenarios are rather uncommon. 2771 2772 Possible values are: 2773 2774 - 0 Forwarding disabled 2775 - 1 Forwarding enabled 2776 2777 **FALSE (0)**: 2778 2779 By default, Host behaviour is assumed. This means: 2780 2781 1. IsRouter flag is not set in Neighbour Advertisements. 2782 2. If accept_ra is TRUE (default), transmit Router 2783 Solicitations. 2784 3. If accept_ra is TRUE (default), accept Router 2785 Advertisements (and do autoconfiguration). 2786 4. If accept_redirects is TRUE (default), accept Redirects. 2787 2788 **TRUE (1)**: 2789 2790 If local forwarding is enabled, Router behaviour is assumed. 2791 This means exactly the reverse from the above: 2792 2793 1. IsRouter flag is set in Neighbour Advertisements. 2794 2. Router Solicitations are not sent unless accept_ra is 2. 2795 3. Router Advertisements are ignored unless accept_ra is 2. 2796 4. Redirects are ignored. 2797 2798 Default: 0 (disabled) if global forwarding is disabled (default), 2799 otherwise 1 (enabled). 2800 2801hop_limit - INTEGER 2802 Default Hop Limit to set. 2803 2804 Default: 64 2805 2806mtu - INTEGER 2807 Default Maximum Transfer Unit 2808 2809 Default: 1280 (IPv6 required minimum) 2810 2811ip_nonlocal_bind - BOOLEAN 2812 If enabled, allows processes to bind() to non-local IPv6 addresses, 2813 which can be quite useful - but may break some applications. 2814 2815 Possible values: 2816 2817 - 0 (disabled) 2818 - 1 (enabled) 2819 2820 Default: 0 (disabled) 2821 2822router_probe_interval - INTEGER 2823 Minimum interval (in seconds) between Router Probing described 2824 in RFC4191. 2825 2826 Default: 60 2827 2828router_solicitation_delay - INTEGER 2829 Number of seconds to wait after interface is brought up 2830 before sending Router Solicitations. 2831 2832 Default: 1 2833 2834router_solicitation_interval - INTEGER 2835 Number of seconds to wait between Router Solicitations. 2836 2837 Default: 4 2838 2839router_solicitations - INTEGER 2840 Number of Router Solicitations to send until assuming no 2841 routers are present. 2842 2843 Default: 3 2844 2845use_oif_addrs_only - BOOLEAN 2846 When enabled, the candidate source addresses for destinations 2847 routed via this interface are restricted to the set of addresses 2848 configured on this interface (vis. RFC 6724, section 4). 2849 2850 Possible values: 2851 2852 - 0 (disabled) 2853 - 1 (enabled) 2854 2855 Default: 0 (disabled) 2856 2857use_tempaddr - INTEGER 2858 Preference for Privacy Extensions (RFC3041). 2859 2860 * <= 0 : disable Privacy Extensions 2861 * == 1 : enable Privacy Extensions, but prefer public 2862 addresses over temporary addresses. 2863 * > 1 : enable Privacy Extensions and prefer temporary 2864 addresses over public addresses. 2865 2866 Default: 2867 2868 * 0 (for most devices) 2869 * -1 (for point-to-point devices and loopback devices) 2870 2871temp_valid_lft - INTEGER 2872 valid lifetime (in seconds) for temporary addresses. If less than the 2873 minimum required lifetime (typically 5-7 seconds), temporary addresses 2874 will not be created. 2875 2876 Default: 172800 (2 days) 2877 2878temp_prefered_lft - INTEGER 2879 Preferred lifetime (in seconds) for temporary addresses. If 2880 temp_prefered_lft is less than the minimum required lifetime (typically 2881 5-7 seconds), the preferred lifetime is the minimum required. If 2882 temp_prefered_lft is greater than temp_valid_lft, the preferred lifetime 2883 is temp_valid_lft. 2884 2885 Default: 86400 (1 day) 2886 2887keep_addr_on_down - INTEGER 2888 Keep all IPv6 addresses on an interface down event. If set static 2889 global addresses with no expiration time are not flushed. 2890 2891 * >0 : enabled 2892 * 0 : system default 2893 * <0 : disabled 2894 2895 Default: 0 (addresses are removed) 2896 2897max_desync_factor - INTEGER 2898 Maximum value for DESYNC_FACTOR, which is a random value 2899 that ensures that clients don't synchronize with each 2900 other and generate new addresses at exactly the same time. 2901 value is in seconds. 2902 2903 Default: 600 2904 2905regen_min_advance - INTEGER 2906 How far in advance (in seconds), at minimum, to create a new temporary 2907 address before the current one is deprecated. This value is added to 2908 the amount of time that may be required for duplicate address detection 2909 to determine when to create a new address. Linux permits setting this 2910 value to less than the default of 2 seconds, but a value less than 2 2911 does not conform to RFC 8981. 2912 2913 Default: 2 2914 2915regen_max_retry - INTEGER 2916 Number of attempts before give up attempting to generate 2917 valid temporary addresses. 2918 2919 Default: 5 2920 2921max_addresses - INTEGER 2922 Maximum number of autoconfigured addresses per interface. Setting 2923 to zero disables the limitation. It is not recommended to set this 2924 value too large (or to zero) because it would be an easy way to 2925 crash the kernel by allowing too many addresses to be created. 2926 2927 Default: 16 2928 2929disable_ipv6 - BOOLEAN 2930 Disable IPv6 operation. If accept_dad is set to 2, this value 2931 will be dynamically set to TRUE if DAD fails for the link-local 2932 address. 2933 2934 Default: FALSE (enable IPv6 operation) 2935 2936 When this value is changed from 1 to 0 (IPv6 is being enabled), 2937 it will dynamically create a link-local address on the given 2938 interface and start Duplicate Address Detection, if necessary. 2939 2940 When this value is changed from 0 to 1 (IPv6 is being disabled), 2941 it will dynamically delete all addresses and routes on the given 2942 interface. From now on it will not possible to add addresses/routes 2943 to the selected interface. 2944 2945accept_dad - INTEGER 2946 Whether to accept DAD (Duplicate Address Detection). 2947 2948 == ============================================================== 2949 0 Disable DAD 2950 1 Enable DAD (default) 2951 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate 2952 link-local address has been found. 2953 == ============================================================== 2954 2955 DAD operation and mode on a given interface will be selected according 2956 to the maximum value of conf/{all,interface}/accept_dad. 2957 2958force_tllao - BOOLEAN 2959 Enable sending the target link-layer address option even when 2960 responding to a unicast neighbor solicitation. 2961 2962 Default: FALSE 2963 2964 Quoting from RFC 2461, section 4.4, Target link-layer address: 2965 2966 "The option MUST be included for multicast solicitations in order to 2967 avoid infinite Neighbor Solicitation "recursion" when the peer node 2968 does not have a cache entry to return a Neighbor Advertisements 2969 message. When responding to unicast solicitations, the option can be 2970 omitted since the sender of the solicitation has the correct link- 2971 layer address; otherwise it would not have be able to send the unicast 2972 solicitation in the first place. However, including the link-layer 2973 address in this case adds little overhead and eliminates a potential 2974 race condition where the sender deletes the cached link-layer address 2975 prior to receiving a response to a previous solicitation." 2976 2977ndisc_notify - BOOLEAN 2978 Define mode for notification of address and device changes. 2979 2980 Possible values: 2981 2982 - 0 (disabled) - do nothing 2983 - 1 (enabled) - Generate unsolicited neighbour advertisements when device is brought 2984 up or hardware address changes. 2985 2986 Default: 0 (disabled) 2987 2988ndisc_tclass - INTEGER 2989 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor 2990 Discovery (Router Solicitation, Router Advertisement, Neighbor 2991 Solicitation, Neighbor Advertisement, Redirect) messages. 2992 These 8 bits can be interpreted as 6 high order bits holding the DSCP 2993 value and 2 low order bits representing ECN (which you probably want 2994 to leave cleared). 2995 2996 * 0 - (default) 2997 2998ndisc_evict_nocarrier - BOOLEAN 2999 Clears the neighbor discovery table on NOCARRIER events. This option is 3000 important for wireless devices where the neighbor discovery cache should 3001 not be cleared when roaming between access points on the same network. 3002 In most cases this should remain as the default (1). 3003 3004 Possible values: 3005 3006 - 0 (disabled) - Do not clear neighbor discovery cache on NOCARRIER events. 3007 - 1 (enabled) - Clear neighbor discover cache on NOCARRIER events. 3008 3009 Default: 1 (enabled) 3010 3011mldv1_unsolicited_report_interval - INTEGER 3012 The interval in milliseconds in which the next unsolicited 3013 MLDv1 report retransmit will take place. 3014 3015 Default: 10000 (10 seconds) 3016 3017mldv2_unsolicited_report_interval - INTEGER 3018 The interval in milliseconds in which the next unsolicited 3019 MLDv2 report retransmit will take place. 3020 3021 Default: 1000 (1 second) 3022 3023force_mld_version - INTEGER 3024 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed 3025 * 1 - Enforce to use MLD version 1 3026 * 2 - Enforce to use MLD version 2 3027 3028suppress_frag_ndisc - INTEGER 3029 Control RFC 6980 (Security Implications of IPv6 Fragmentation 3030 with IPv6 Neighbor Discovery) behavior: 3031 3032 * 1 - (default) discard fragmented neighbor discovery packets 3033 * 0 - allow fragmented neighbor discovery packets 3034 3035optimistic_dad - BOOLEAN 3036 Whether to perform Optimistic Duplicate Address Detection (RFC 4429). 3037 3038 Optimistic Duplicate Address Detection for the interface will be enabled 3039 if at least one of conf/{all,interface}/optimistic_dad is set to 1, 3040 it will be disabled otherwise. 3041 3042 Possible values: 3043 3044 - 0 (disabled) 3045 - 1 (enabled) 3046 3047 Default: 0 (disabled) 3048 3049 3050use_optimistic - BOOLEAN 3051 If enabled, do not classify optimistic addresses as deprecated during 3052 source address selection. Preferred addresses will still be chosen 3053 before optimistic addresses, subject to other ranking in the source 3054 address selection algorithm. 3055 3056 This will be enabled if at least one of 3057 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise. 3058 3059 Possible values: 3060 3061 - 0 (disabled) 3062 - 1 (enabled) 3063 3064 Default: 0 (disabled) 3065 3066stable_secret - IPv6 address 3067 This IPv6 address will be used as a secret to generate IPv6 3068 addresses for link-local addresses and autoconfigured 3069 ones. All addresses generated after setting this secret will 3070 be stable privacy ones by default. This can be changed via the 3071 addrgenmode ip-link. conf/default/stable_secret is used as the 3072 secret for the namespace, the interface specific ones can 3073 overwrite that. Writes to conf/all/stable_secret are refused. 3074 3075 It is recommended to generate this secret during installation 3076 of a system and keep it stable after that. 3077 3078 By default the stable secret is unset. 3079 3080addr_gen_mode - INTEGER 3081 Defines how link-local and autoconf addresses are generated. 3082 3083 = ================================================================= 3084 0 generate address based on EUI64 (default) 3085 1 do no generate a link-local address, use EUI64 for addresses 3086 generated from autoconf 3087 2 generate stable privacy addresses, using the secret from 3088 stable_secret (RFC7217) 3089 3 generate stable privacy addresses, using a random secret if unset 3090 = ================================================================= 3091 3092drop_unicast_in_l2_multicast - BOOLEAN 3093 Drop any unicast IPv6 packets that are received in link-layer 3094 multicast (or broadcast) frames. 3095 3096 Possible values: 3097 3098 - 0 (disabled) 3099 - 1 (enabled) 3100 3101 Default: 0 (disabled) 3102 3103drop_unsolicited_na - BOOLEAN 3104 Drop all unsolicited neighbor advertisements, for example if there's 3105 a known good NA proxy on the network and such frames need not be used 3106 (or in the case of 802.11, must not be used to prevent attacks.) 3107 3108 Possible values: 3109 3110 - 0 (disabled) 3111 - 1 (enabled) 3112 3113 Default: 0 (disabled). 3114 3115accept_untracked_na - INTEGER 3116 Define behavior for accepting neighbor advertisements from devices that 3117 are absent in the neighbor cache: 3118 3119 - 0 - (default) Do not accept unsolicited and untracked neighbor 3120 advertisements. 3121 3122 - 1 - Add a new neighbor cache entry in STALE state for routers on 3123 receiving a neighbor advertisement (either solicited or unsolicited) 3124 with target link-layer address option specified if no neighbor entry 3125 is already present for the advertised IPv6 address. Without this knob, 3126 NAs received for untracked addresses (absent in neighbor cache) are 3127 silently ignored. 3128 3129 This is as per router-side behavior documented in RFC9131. 3130 3131 This has lower precedence than drop_unsolicited_na. 3132 3133 This will optimize the return path for the initial off-link 3134 communication that is initiated by a directly connected host, by 3135 ensuring that the first-hop router which turns on this setting doesn't 3136 have to buffer the initial return packets to do neighbor-solicitation. 3137 The prerequisite is that the host is configured to send unsolicited 3138 neighbor advertisements on interface bringup. This setting should be 3139 used in conjunction with the ndisc_notify setting on the host to 3140 satisfy this prerequisite. 3141 3142 - 2 - Extend option (1) to add a new neighbor cache entry only if the 3143 source IP address is in the same subnet as an address configured on 3144 the interface that received the neighbor advertisement. 3145 3146enhanced_dad - BOOLEAN 3147 Include a nonce option in the IPv6 neighbor solicitation messages used for 3148 duplicate address detection per RFC7527. A received DAD NS will only signal 3149 a duplicate address if the nonce is different. This avoids any false 3150 detection of duplicates due to loopback of the NS messages that we send. 3151 The nonce option will be sent on an interface unless both of 3152 conf/{all,interface}/enhanced_dad are set to FALSE. 3153 3154 Possible values: 3155 3156 - 0 (disabled) 3157 - 1 (enabled) 3158 3159 Default: 1 (enabled) 3160 3161``icmp/*``: 3162=========== 3163 3164ratelimit - INTEGER 3165 Limit the maximal rates for sending ICMPv6 messages. 3166 3167 0 to disable any limiting, 3168 otherwise the minimal space between responses in milliseconds. 3169 3170 Default: 1000 3171 3172ratemask - list of comma separated ranges 3173 For ICMPv6 message types matching the ranges in the ratemask, limit 3174 the sending of the message according to ratelimit parameter. 3175 3176 The format used for both input and output is a comma separated 3177 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and 3178 129). Writing to the file will clear all previous ranges of ICMPv6 3179 message types and update the current list with the input. 3180 3181 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml 3182 for numerical values of ICMPv6 message types, e.g. echo request is 128 3183 and echo reply is 129. 3184 3185 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big) 3186 3187echo_ignore_all - BOOLEAN 3188 If enabled, then the kernel will ignore all ICMP ECHO 3189 requests sent to it over the IPv6 protocol. 3190 3191 Possible values: 3192 3193 - 0 (disabled) 3194 - 1 (enabled) 3195 3196 Default: 0 (disabled) 3197 3198echo_ignore_multicast - BOOLEAN 3199 If enabled, then the kernel will ignore all ICMP ECHO 3200 requests sent to it over the IPv6 protocol via multicast. 3201 3202 Possible values: 3203 3204 - 0 (disabled) 3205 - 1 (enabled) 3206 3207 Default: 0 (disabled) 3208 3209echo_ignore_anycast - BOOLEAN 3210 If enabled, then the kernel will ignore all ICMP ECHO 3211 requests sent to it over the IPv6 protocol destined to anycast address. 3212 3213 Possible values: 3214 3215 - 0 (disabled) 3216 - 1 (enabled) 3217 3218 Default: 0 (disabled) 3219 3220error_anycast_as_unicast - BOOLEAN 3221 If enabled, then the kernel will respond with ICMP Errors 3222 resulting from requests sent to it over the IPv6 protocol destined 3223 to anycast address essentially treating anycast as unicast. 3224 3225 Possible values: 3226 3227 - 0 (disabled) 3228 - 1 (enabled) 3229 3230 Default: 0 (disabled) 3231 3232xfrm6_gc_thresh - INTEGER 3233 (Obsolete since linux-4.14) 3234 The threshold at which we will start garbage collecting for IPv6 3235 destination cache entries. At twice this value the system will 3236 refuse new allocations. 3237 3238 3239IPv6 Update by: 3240Pekka Savola <pekkas@netcore.fi> 3241YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org> 3242 3243 3244/proc/sys/net/bridge/* Variables: 3245================================= 3246 3247bridge-nf-call-arptables - BOOLEAN 3248 3249 Possible values: 3250 3251 - 0 (disabled) - disable this. 3252 - 1 (enabled) - pass bridged ARP traffic to arptables' FORWARD chain. 3253 3254 Default: 1 (enabled) 3255 3256bridge-nf-call-iptables - BOOLEAN 3257 3258 Possible values: 3259 3260 - 0 (disabled) - disable this. 3261 - 1 (enabled) - pass bridged IPv4 traffic to iptables' chains. 3262 3263 Default: 1 (enabled) 3264 3265bridge-nf-call-ip6tables - BOOLEAN 3266 3267 Possible values: 3268 3269 - 0 (disabled) - disable this. 3270 - 1 (enabled) - pass bridged IPv6 traffic to ip6tables' chains. 3271 3272 Default: 1 (enabled) 3273 3274bridge-nf-filter-vlan-tagged - BOOLEAN 3275 3276 Possible values: 3277 3278 - 0 (disabled) - disable this. 3279 - 1 (enabled) - pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables 3280 3281 Default: 0 (disabled) 3282 3283bridge-nf-filter-pppoe-tagged - BOOLEAN 3284 3285 Possible values: 3286 3287 - 0 (disabled) - disable this. 3288 - 1 (enabled) - pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables. 3289 3290 Default: 0 (disabled) 3291 3292bridge-nf-pass-vlan-input-dev - BOOLEAN 3293 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan 3294 interface on the bridge and set the netfilter input device to the 3295 vlan. This allows use of e.g. "iptables -i br0.1" and makes the 3296 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no 3297 matching vlan interface is found, or this switch is off, the input 3298 device is set to the bridge interface. 3299 3300 - 0: disable bridge netfilter vlan interface lookup. 3301 3302 Default: 0 3303 3304``proc/sys/net/sctp/*`` Variables: 3305================================== 3306 3307addip_enable - BOOLEAN 3308 Enable or disable extension of Dynamic Address Reconfiguration 3309 (ADD-IP) functionality specified in RFC5061. This extension provides 3310 the ability to dynamically add and remove new addresses for the SCTP 3311 associations. 3312 3313 Possible values: 3314 3315 - 0 (disabled) - disable extension. 3316 - 1 (enabled) - enable extension 3317 3318 Default: 0 (disabled) 3319 3320pf_enable - INTEGER 3321 Enable or disable pf (pf is short for potentially failed) state. A value 3322 of pf_retrans > path_max_retrans also disables pf state. That is, one of 3323 both pf_enable and pf_retrans > path_max_retrans can disable pf state. 3324 Since pf_retrans and path_max_retrans can be changed by userspace 3325 application, sometimes user expects to disable pf state by the value of 3326 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans 3327 or path_max_retrans is changed by the user application, this pf state is 3328 enabled. As such, it is necessary to add this to dynamically enable 3329 and disable pf state. See: 3330 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for 3331 details. 3332 3333 Possible values: 3334 3335 - 1: Enable pf. 3336 - 0: Disable pf. 3337 3338 Default: 1 3339 3340pf_expose - INTEGER 3341 Unset or enable/disable pf (pf is short for potentially failed) state 3342 exposure. Applications can control the exposure of the PF path state 3343 in the SCTP_PEER_ADDR_CHANGE event and access of SCTP_PF-state 3344 transport info via SCTP_GET_PEER_ADDR_INFO sockopt. 3345 3346 Possible values: 3347 3348 - 0: Unset pf state exposure (compatible with old applications). No 3349 event will be sent but the transport info can be queried. 3350 - 1: Disable pf state exposure. No event will be sent and trying to 3351 obtain transport info will return -EACCESS. 3352 - 2: Enable pf state exposure. The event will be sent for a transport 3353 becoming SCTP_PF state and transport info can be obtained. 3354 3355 Default: 0 3356 3357addip_noauth_enable - BOOLEAN 3358 Dynamic Address Reconfiguration (ADD-IP) requires the use of 3359 authentication to protect the operations of adding or removing new 3360 addresses. This requirement is mandated so that unauthorized hosts 3361 would not be able to hijack associations. However, older 3362 implementations may not have implemented this requirement while 3363 allowing the ADD-IP extension. For reasons of interoperability, 3364 we provide this variable to control the enforcement of the 3365 authentication requirement. 3366 3367 == =============================================================== 3368 1 Allow ADD-IP extension to be used without authentication. This 3369 should only be set in a closed environment for interoperability 3370 with older implementations. 3371 3372 0 Enforce the authentication requirement 3373 == =============================================================== 3374 3375 Default: 0 3376 3377auth_enable - BOOLEAN 3378 Enable or disable Authenticated Chunks extension. This extension 3379 provides the ability to send and receive authenticated chunks and is 3380 required for secure operation of Dynamic Address Reconfiguration 3381 (ADD-IP) extension. 3382 3383 Possible values: 3384 3385 - 0 (disabled) - disable extension. 3386 - 1 (enabled) - enable extension 3387 3388 Default: 0 (disabled) 3389 3390prsctp_enable - BOOLEAN 3391 Enable or disable the Partial Reliability extension (RFC3758) which 3392 is used to notify peers that a given DATA should no longer be expected. 3393 3394 Possible values: 3395 3396 - 0 (disabled) - disable extension. 3397 - 1 (enabled) - enable extension 3398 3399 Default: 1 (enabled) 3400 3401max_burst - INTEGER 3402 The limit of the number of new packets that can be initially sent. It 3403 controls how bursty the generated traffic can be. 3404 3405 Default: 4 3406 3407association_max_retrans - INTEGER 3408 Set the maximum number for retransmissions that an association can 3409 attempt deciding that the remote end is unreachable. If this value 3410 is exceeded, the association is terminated. 3411 3412 Default: 10 3413 3414max_init_retransmits - INTEGER 3415 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks 3416 that an association will attempt before declaring the destination 3417 unreachable and terminating. 3418 3419 Default: 8 3420 3421path_max_retrans - INTEGER 3422 The maximum number of retransmissions that will be attempted on a given 3423 path. Once this threshold is exceeded, the path is considered 3424 unreachable, and new traffic will use a different path when the 3425 association is multihomed. 3426 3427 Default: 5 3428 3429pf_retrans - INTEGER 3430 The number of retransmissions that will be attempted on a given path 3431 before traffic is redirected to an alternate transport (should one 3432 exist). Note this is distinct from path_max_retrans, as a path that 3433 passes the pf_retrans threshold can still be used. Its only 3434 deprioritized when a transmission path is selected by the stack. This 3435 setting is primarily used to enable fast failover mechanisms without 3436 having to reduce path_max_retrans to a very low value. See: 3437 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt 3438 for details. Note also that a value of pf_retrans > path_max_retrans 3439 disables this feature. Since both pf_retrans and path_max_retrans can 3440 be changed by userspace application, a variable pf_enable is used to 3441 disable pf state. 3442 3443 Default: 0 3444 3445ps_retrans - INTEGER 3446 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming 3447 from section-5 "Primary Path Switchover" in rfc7829. The primary path 3448 will be changed to another active path when the path error counter on 3449 the old primary path exceeds PSMR, so that "the SCTP sender is allowed 3450 to continue data transmission on a new working path even when the old 3451 primary destination address becomes active again". Note this feature 3452 is disabled by initializing 'ps_retrans' per netns as 0xffff by default, 3453 and its value can't be less than 'pf_retrans' when changing by sysctl. 3454 3455 Default: 0xffff 3456 3457rto_initial - INTEGER 3458 The initial round trip timeout value in milliseconds that will be used 3459 in calculating round trip times. This is the initial time interval 3460 for retransmissions. 3461 3462 Default: 3000 3463 3464rto_max - INTEGER 3465 The maximum value (in milliseconds) of the round trip timeout. This 3466 is the largest time interval that can elapse between retransmissions. 3467 3468 Default: 60000 3469 3470rto_min - INTEGER 3471 The minimum value (in milliseconds) of the round trip timeout. This 3472 is the smallest time interval the can elapse between retransmissions. 3473 3474 Default: 1000 3475 3476hb_interval - INTEGER 3477 The interval (in milliseconds) between HEARTBEAT chunks. These chunks 3478 are sent at the specified interval on idle paths to probe the state of 3479 a given path between 2 associations. 3480 3481 Default: 30000 3482 3483sack_timeout - INTEGER 3484 The amount of time (in milliseconds) that the implementation will wait 3485 to send a SACK. 3486 3487 Default: 200 3488 3489valid_cookie_life - INTEGER 3490 The default lifetime of the SCTP cookie (in milliseconds). The cookie 3491 is used during association establishment. 3492 3493 Default: 60000 3494 3495cookie_preserve_enable - BOOLEAN 3496 Enable or disable the ability to extend the lifetime of the SCTP cookie 3497 that is used during the establishment phase of SCTP association 3498 3499 Possible values: 3500 3501 - 0 (disabled) - disable. 3502 - 1 (enabled) - enable cookie lifetime extension. 3503 3504 Default: 1 (enabled) 3505 3506cookie_hmac_alg - STRING 3507 Select the hmac algorithm used when generating the cookie value sent by 3508 a listening sctp socket to a connecting client in the INIT-ACK chunk. 3509 Valid values are: 3510 3511 * md5 3512 * sha1 3513 * none 3514 3515 Ability to assign md5 or sha1 as the selected alg is predicated on the 3516 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and 3517 CONFIG_CRYPTO_SHA1). 3518 3519 Default: Dependent on configuration. MD5 if available, else SHA1 if 3520 available, else none. 3521 3522rcvbuf_policy - INTEGER 3523 Determines if the receive buffer is attributed to the socket or to 3524 association. SCTP supports the capability to create multiple 3525 associations on a single socket. When using this capability, it is 3526 possible that a single stalled association that's buffering a lot 3527 of data may block other associations from delivering their data by 3528 consuming all of the receive buffer space. To work around this, 3529 the rcvbuf_policy could be set to attribute the receiver buffer space 3530 to each association instead of the socket. This prevents the described 3531 blocking. 3532 3533 - 1: rcvbuf space is per association 3534 - 0: rcvbuf space is per socket 3535 3536 Default: 0 3537 3538sndbuf_policy - INTEGER 3539 Similar to rcvbuf_policy above, this applies to send buffer space. 3540 3541 - 1: Send buffer is tracked per association 3542 - 0: Send buffer is tracked per socket. 3543 3544 Default: 0 3545 3546sctp_mem - vector of 3 INTEGERs: min, pressure, max 3547 Number of pages allowed for queueing by all SCTP sockets. 3548 3549 * min: Below this number of pages SCTP is not bothered about its 3550 memory usage. When amount of memory allocated by SCTP exceeds 3551 this number, SCTP starts to moderate memory usage. 3552 * pressure: This value was introduced to follow format of tcp_mem. 3553 * max: Maximum number of allowed pages. 3554 3555 Default is calculated at boot time from amount of available memory. 3556 3557sctp_rmem - vector of 3 INTEGERs: min, default, max 3558 Only the first value ("min") is used, "default" and "max" are 3559 ignored. 3560 3561 * min: Minimal size of receive buffer used by SCTP socket. 3562 It is guaranteed to each SCTP socket (but not association) even 3563 under moderate memory pressure. 3564 3565 Default: 4K 3566 3567sctp_wmem - vector of 3 INTEGERs: min, default, max 3568 Only the first value ("min") is used, "default" and "max" are 3569 ignored. 3570 3571 * min: Minimum size of send buffer that can be used by SCTP sockets. 3572 It is guaranteed to each SCTP socket (but not association) even 3573 under moderate memory pressure. 3574 3575 Default: 4K 3576 3577addr_scope_policy - INTEGER 3578 Control IPv4 address scoping (see 3579 https://datatracker.ietf.org/doc/draft-stewart-tsvwg-sctp-ipv4/00/ 3580 for details). 3581 3582 - 0 - Disable IPv4 address scoping 3583 - 1 - Enable IPv4 address scoping 3584 - 2 - Follow draft but allow IPv4 private addresses 3585 - 3 - Follow draft but allow IPv4 link local addresses 3586 3587 Default: 1 3588 3589udp_port - INTEGER 3590 The listening port for the local UDP tunneling sock. Normally it's 3591 using the IANA-assigned UDP port number 9899 (sctp-tunneling). 3592 3593 This UDP sock is used for processing the incoming UDP-encapsulated 3594 SCTP packets (from RFC6951), and shared by all applications in the 3595 same net namespace. This UDP sock will be closed when the value is 3596 set to 0. 3597 3598 The value will also be used to set the src port of the UDP header 3599 for the outgoing UDP-encapsulated SCTP packets. For the dest port, 3600 please refer to 'encap_port' below. 3601 3602 Default: 0 3603 3604encap_port - INTEGER 3605 The default remote UDP encapsulation port. 3606 3607 This value is used to set the dest port of the UDP header for the 3608 outgoing UDP-encapsulated SCTP packets by default. Users can also 3609 change the value for each sock/asoc/transport by using setsockopt. 3610 For further information, please refer to RFC6951. 3611 3612 Note that when connecting to a remote server, the client should set 3613 this to the port that the UDP tunneling sock on the peer server is 3614 listening to and the local UDP tunneling sock on the client also 3615 must be started. On the server, it would get the encap_port from 3616 the incoming packet's source port. 3617 3618 Default: 0 3619 3620plpmtud_probe_interval - INTEGER 3621 The time interval (in milliseconds) for the PLPMTUD probe timer, 3622 which is configured to expire after this period to receive an 3623 acknowledgment to a probe packet. This is also the time interval 3624 between the probes for the current pmtu when the probe search 3625 is done. 3626 3627 PLPMTUD will be disabled when 0 is set, and other values for it 3628 must be >= 5000. 3629 3630 Default: 0 3631 3632reconf_enable - BOOLEAN 3633 Enable or disable extension of Stream Reconfiguration functionality 3634 specified in RFC6525. This extension provides the ability to "reset" 3635 a stream, and it includes the Parameters of "Outgoing/Incoming SSN 3636 Reset", "SSN/TSN Reset" and "Add Outgoing/Incoming Streams". 3637 3638 Possible values: 3639 3640 - 0 (disabled) - Disable extension. 3641 - 1 (enabled) - Enable extension. 3642 3643 Default: 0 (disabled) 3644 3645intl_enable - BOOLEAN 3646 Enable or disable extension of User Message Interleaving functionality 3647 specified in RFC8260. This extension allows the interleaving of user 3648 messages sent on different streams. With this feature enabled, I-DATA 3649 chunk will replace DATA chunk to carry user messages if also supported 3650 by the peer. Note that to use this feature, one needs to set this option 3651 to 1 and also needs to set socket options SCTP_FRAGMENT_INTERLEAVE to 2 3652 and SCTP_INTERLEAVING_SUPPORTED to 1. 3653 3654 Possible values: 3655 3656 - 0 (disabled) - Disable extension. 3657 - 1 (enabled) - Enable extension. 3658 3659 Default: 0 (disabled) 3660 3661ecn_enable - BOOLEAN 3662 Control use of Explicit Congestion Notification (ECN) by SCTP. 3663 Like in TCP, ECN is used only when both ends of the SCTP connection 3664 indicate support for it. This feature is useful in avoiding losses 3665 due to congestion by allowing supporting routers to signal congestion 3666 before having to drop packets. 3667 3668 Possible values: 3669 3670 - 0 (disabled) - Disable ecn. 3671 - 1 (enabled) - Enable ecn. 3672 3673 Default: 1 (enabled) 3674 3675l3mdev_accept - BOOLEAN 3676 Enabling this option allows a "global" bound socket to work 3677 across L3 master domains (e.g., VRFs) with packets capable of 3678 being received regardless of the L3 domain in which they 3679 originated. Only valid when the kernel was compiled with 3680 CONFIG_NET_L3_MASTER_DEV. 3681 3682 Possible values: 3683 3684 - 0 (disabled) 3685 - 1 (enabled) 3686 3687 Default: 1 (enabled) 3688 3689 3690``/proc/sys/net/core/*`` 3691======================== 3692 3693 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries. 3694 3695 3696``/proc/sys/net/unix/*`` 3697======================== 3698 3699max_dgram_qlen - INTEGER 3700 The maximum length of dgram socket receive queue 3701 3702 Default: 10 3703 3704