1 .. SPDX-License-Identifier: GPL-2.0
7 /proc/sys/net/ipv4/* Variables
8 ==============================
11 - 0 - disabled (default)
14 Forward Packets between interfaces.
16 This variable is special, its change resets all configuration
17 parameters to their default state (RFC1122 for hosts, RFC1812
20 ip_default_ttl - INTEGER
21 Default value of TTL field (Time To Live) for outgoing (but not
22 forwarded) IP packets. Should be between 1 and 255 inclusive.
23 Default: 64 (as recommended by RFC1700)
25 ip_no_pmtu_disc - INTEGER
26 Disable Path MTU Discovery. If enabled in mode 1 and a
27 fragmentation-required ICMP is received, the PMTU to this
28 destination will be set to min_pmtu (see below). You will need
29 to raise min_pmtu to the smallest interface MTU on your system
30 manually if you want to avoid locally generated fragments.
32 In mode 2 incoming Path MTU Discovery messages will be
33 discarded. Outgoing frames are handled the same as in mode 1,
34 implicitly setting IP_PMTUDISC_DONT on every created socket.
36 Mode 3 is a hardened pmtu discover mode. The kernel will only
37 accept fragmentation-needed errors if the underlying protocol
38 can verify them besides a plain socket lookup. Current
39 protocols for which pmtu events will be honored are TCP, SCTP
40 and DCCP as they verify e.g. the sequence number or the
41 association. This mode should not be enabled globally but is
42 only intended to secure e.g. name servers in namespaces where
43 TCP path mtu must still work but path MTU information of other
44 protocols should be discarded. If enabled globally this mode
45 could break other protocols.
52 default 552 - minimum discovered Path MTU
54 ip_forward_use_pmtu - BOOLEAN
55 By default we don't trust protocol path MTUs while forwarding
56 because they could be easily forged and can lead to unwanted
57 fragmentation by the router.
58 You only need to enable this if you have user-space software
59 which tries to discover path mtus by itself and depends on the
60 kernel honoring this information. This is normally not the
70 fwmark_reflect - BOOLEAN
71 Controls the fwmark of kernel-generated IPv4 reply packets that are not
72 associated with a socket for example, TCP RSTs or ICMP echo replies).
73 If unset, these packets have a fwmark of zero. If set, they have the
74 fwmark of the packet they are replying to.
78 fib_multipath_use_neigh - BOOLEAN
79 Use status of existing neighbor entry when determining nexthop for
80 multipath routes. If disabled, neighbor information is not used and
81 packets could be directed to a failed nexthop. Only valid for kernels
82 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
91 fib_multipath_hash_policy - INTEGER
92 Controls which hash policy to use for multipath routes. Only valid
93 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
101 - 2 - Layer 3 or inner Layer 3 if present
102 - 3 - Custom multipath hash. Fields used for multipath hash calculation
103 are determined by fib_multipath_hash_fields sysctl
105 fib_multipath_hash_fields - UNSIGNED INTEGER
106 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
107 fields used for multipath hash calculation are determined by this
110 This value is a bitmask which enables various fields for multipath hash
115 ====== ============================
116 0x0001 Source IP address
117 0x0002 Destination IP address
119 0x0008 Unused (Flow Label)
121 0x0020 Destination port
122 0x0040 Inner source IP address
123 0x0080 Inner destination IP address
124 0x0100 Inner IP protocol
125 0x0200 Inner Flow Label
126 0x0400 Inner source port
127 0x0800 Inner destination port
128 ====== ============================
130 Default: 0x0007 (source IP, destination IP and IP protocol)
132 fib_sync_mem - UNSIGNED INTEGER
133 Amount of dirty memory from fib entries that can be backlogged before
134 synchronize_rcu is forced.
136 Default: 512kB Minimum: 64kB Maximum: 64MB
138 ip_forward_update_priority - INTEGER
139 Whether to update SKB priority from "TOS" field in IPv4 header after it
140 is forwarded. The new SKB priority is mapped from TOS field value
141 according to an rt_tos2priority table (see e.g. man tc-prio).
143 Default: 1 (Update priority.)
147 - 0 - Do not update priority.
148 - 1 - Update priority.
150 route/max_size - INTEGER
151 Maximum number of routes allowed in the kernel. Increase
152 this when using large numbers of interfaces and/or routes.
154 From linux kernel 3.6 onwards, this is deprecated for ipv4
155 as route cache is no longer used.
157 neigh/default/gc_thresh1 - INTEGER
158 Minimum number of entries to keep. Garbage collector will not
159 purge entries if there are fewer than this number.
163 neigh/default/gc_thresh2 - INTEGER
164 Threshold when garbage collector becomes more aggressive about
165 purging entries. Entries older than 5 seconds will be cleared
166 when over this number.
170 neigh/default/gc_thresh3 - INTEGER
171 Maximum number of non-PERMANENT neighbor entries allowed. Increase
172 this when using large numbers of interfaces and when communicating
173 with large numbers of directly-connected peers.
177 neigh/default/unres_qlen_bytes - INTEGER
178 The maximum number of bytes which may be used by packets
179 queued for each unresolved address by other network layers.
182 Setting negative value is meaningless and will return error.
184 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
186 Exact value depends on architecture and kernel options,
187 but should be enough to allow queuing 256 packets
190 neigh/default/unres_qlen - INTEGER
191 The maximum number of packets which may be queued for each
192 unresolved address by other network layers.
194 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
196 Prior to linux 3.3, the default value is 3 which may cause
197 unexpected packet loss. The current default value is calculated
198 according to default value of unres_qlen_bytes and true size of
203 mtu_expires - INTEGER
204 Time, in seconds, that cached PMTU information is kept.
206 min_adv_mss - INTEGER
207 The advertised MSS depends on the first hop route MTU, but will
208 never be lower than this setting.
210 fib_notify_on_flag_change - INTEGER
211 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
212 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
214 After installing a route to the kernel, user space receives an
215 acknowledgment, which means the route was installed in the kernel,
216 but not necessarily in hardware.
217 It is also possible for a route already installed in hardware to change
218 its action and therefore its flags. For example, a host route that is
219 trapping packets can be "promoted" to perform decapsulation following
220 the installation of an IPinIP/VXLAN tunnel.
221 The notifications will indicate to user-space the state of the route.
223 Default: 0 (Do not emit notifications.)
227 - 0 - Do not emit notifications.
228 - 1 - Emit notifications.
229 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
233 ipfrag_high_thresh - LONG INTEGER
234 Maximum memory used to reassemble IP fragments.
236 ipfrag_low_thresh - LONG INTEGER
237 (Obsolete since linux-4.17)
238 Maximum memory used to reassemble IP fragments before the kernel
239 begins to remove incomplete fragment queues to free up resources.
240 The kernel still accepts new fragments for defragmentation.
242 ipfrag_time - INTEGER
243 Time in seconds to keep an IP fragment in memory.
245 ipfrag_max_dist - INTEGER
246 ipfrag_max_dist is a non-negative integer value which defines the
247 maximum "disorder" which is allowed among fragments which share a
248 common IP source address. Note that reordering of packets is
249 not unusual, but if a large number of fragments arrive from a source
250 IP address while a particular fragment queue remains incomplete, it
251 probably indicates that one or more fragments belonging to that queue
252 have been lost. When ipfrag_max_dist is positive, an additional check
253 is done on fragments before they are added to a reassembly queue - if
254 ipfrag_max_dist (or more) fragments have arrived from a particular IP
255 address between additions to any IP fragment queue using that source
256 address, it's presumed that one or more fragments in the queue are
257 lost. The existing fragment queue will be dropped, and a new one
258 started. An ipfrag_max_dist value of zero disables this check.
260 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
261 result in unnecessarily dropping fragment queues when normal
262 reordering of packets occurs, which could lead to poor application
263 performance. Using a very large value, e.g. 50000, increases the
264 likelihood of incorrectly reassembling IP fragments that originate
265 from different IP datagrams, which could result in data corruption.
271 inet_peer_threshold - INTEGER
272 The approximate size of the storage. Starting from this threshold
273 entries will be thrown aggressively. This threshold also determines
274 entries' time-to-live and time intervals between garbage collection
275 passes. More entries, less time-to-live, less GC interval.
277 inet_peer_minttl - INTEGER
278 Minimum time-to-live of entries. Should be enough to cover fragment
279 time-to-live on the reassembling side. This minimum time-to-live is
280 guaranteed if the pool size is less than inet_peer_threshold.
283 inet_peer_maxttl - INTEGER
284 Maximum time-to-live of entries. Unused entries will expire after
285 this period of time if there is no memory pressure on the pool (i.e.
286 when the number of entries in the pool is very small).
293 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
294 Defaults to 4096. (Was 128 before linux-5.4)
295 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
297 tcp_abort_on_overflow - BOOLEAN
298 If listening service is too slow to accept new connections,
299 reset them. Default state is FALSE. It means that if overflow
300 occurred due to a burst, connection will recover. Enable this
301 option _only_ if you are really sure that listening daemon
302 cannot be tuned to accept connections faster. Enabling this
303 option can harm clients of your server.
305 tcp_adv_win_scale - INTEGER
306 Count buffering overhead as bytes/2^tcp_adv_win_scale
307 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
310 Possible values are [-31, 31], inclusive.
314 tcp_allowed_congestion_control - STRING
315 Show/set the congestion control choices available to non-privileged
316 processes. The list is a subset of those listed in
317 tcp_available_congestion_control.
319 Default is "reno" and the default setting (tcp_congestion_control).
321 tcp_app_win - INTEGER
322 Reserve max(window/2^tcp_app_win, mss) of window for application
323 buffer. Value 0 is special, it means that nothing is reserved.
325 Possible values are [0, 31], inclusive.
329 tcp_autocorking - BOOLEAN
330 Enable TCP auto corking :
331 When applications do consecutive small write()/sendmsg() system calls,
332 we try to coalesce these small writes as much as possible, to lower
333 total amount of sent packets. This is done if at least one prior
334 packet for the flow is waiting in Qdisc queues or device transmit
335 queue. Applications can still use TCP_CORK for optimal behavior
336 when they know how/when to uncork their sockets.
340 tcp_available_congestion_control - STRING
341 Shows the available congestion control choices that are registered.
342 More congestion control algorithms may be available as modules,
345 tcp_base_mss - INTEGER
346 The initial value of search_low to be used by the packetization layer
347 Path MTU discovery (MTU probing). If MTU probing is enabled,
348 this is the initial MSS used by the connection.
350 tcp_mtu_probe_floor - INTEGER
351 If MTU probing is enabled this caps the minimum MSS used for search_low
356 tcp_min_snd_mss - INTEGER
357 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
358 as described in RFC 1122 and RFC 6691.
360 If this ADVMSS option is smaller than tcp_min_snd_mss,
361 it is silently capped to tcp_min_snd_mss.
363 Default : 48 (at least 8 bytes of payload per segment)
365 tcp_congestion_control - STRING
366 Set the congestion control algorithm to be used for new
367 connections. The algorithm "reno" is always available, but
368 additional choices may be available based on kernel configuration.
369 Default is set as part of kernel configuration.
370 For passive connections, the listener congestion control choice
373 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
376 Allows TCP to send "duplicate" SACKs.
378 tcp_early_retrans - INTEGER
379 Tail loss probe (TLP) converts RTOs occurring due to tail
380 losses into fast recovery (draft-ietf-tcpm-rack). Note that
381 TLP requires RACK to function properly (see tcp_recovery below)
391 Control use of Explicit Congestion Notification (ECN) by TCP.
392 ECN is used only when both ends of the TCP connection indicate
393 support for it. This feature is useful in avoiding losses due
394 to congestion by allowing supporting routers to signal
395 congestion before having to drop packets.
399 = =====================================================
400 0 Disable ECN. Neither initiate nor accept ECN.
401 1 Enable ECN when requested by incoming connections and
402 also request ECN on outgoing connection attempts.
403 2 Enable ECN when requested by incoming connections
404 but do not request ECN on outgoing connections.
405 = =====================================================
409 tcp_ecn_fallback - BOOLEAN
410 If the kernel detects that ECN connection misbehaves, enable fall
411 back to non-ECN. Currently, this knob implements the fallback
412 from RFC3168, section 6.1.1.1., but we reserve that in future,
413 additional detection mechanisms could be implemented under this
414 knob. The value is not used, if tcp_ecn or per route (or congestion
415 control) ECN settings are disabled.
417 Default: 1 (fallback enabled)
420 This is a legacy option, it has no effect anymore.
422 tcp_fin_timeout - INTEGER
423 The length of time an orphaned (no longer referenced by any
424 application) connection will remain in the FIN_WAIT_2 state
425 before it is aborted at the local end. While a perfectly
426 valid "receive only" state for an un-orphaned connection, an
427 orphaned connection in FIN_WAIT_2 state could otherwise wait
428 forever for the remote to close its end of the connection.
435 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
436 F-RTO is an enhanced recovery algorithm for TCP retransmission
437 timeouts. It is particularly beneficial in networks where the
438 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
439 modification. It does not require any support from the peer.
441 By default it's enabled with a non-zero value. 0 disables F-RTO.
443 tcp_fwmark_accept - BOOLEAN
444 If set, incoming connections to listening sockets that do not have a
445 socket mark will set the mark of the accepting socket to the fwmark of
446 the incoming SYN packet. This will cause all packets on that connection
447 (starting from the first SYNACK) to be sent with that fwmark. The
448 listening socket's mark is unchanged. Listening sockets that already
449 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
454 tcp_invalid_ratelimit - INTEGER
455 Limit the maximal rate for sending duplicate acknowledgments
456 in response to incoming TCP packets that are for an existing
457 connection but that are invalid due to any of these reasons:
459 (a) out-of-window sequence number,
460 (b) out-of-window acknowledgment number, or
461 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
463 This can help mitigate simple "ack loop" DoS attacks, wherein
464 a buggy or malicious middlebox or man-in-the-middle can
465 rewrite TCP header fields in manner that causes each endpoint
466 to think that the other is sending invalid TCP segments, thus
467 causing each side to send an unterminating stream of duplicate
468 acknowledgments for invalid segments.
470 Using 0 disables rate-limiting of dupacks in response to
471 invalid segments; otherwise this value specifies the minimal
472 space between sending such dupacks, in milliseconds.
474 Default: 500 (milliseconds).
476 tcp_keepalive_time - INTEGER
477 How often TCP sends out keepalive messages when keepalive is enabled.
480 tcp_keepalive_probes - INTEGER
481 How many keepalive probes TCP sends out, until it decides that the
482 connection is broken. Default value: 9.
484 tcp_keepalive_intvl - INTEGER
485 How frequently the probes are send out. Multiplied by
486 tcp_keepalive_probes it is time to kill not responding connection,
487 after probes started. Default value: 75sec i.e. connection
488 will be aborted after ~11 minutes of retries.
490 tcp_l3mdev_accept - BOOLEAN
491 Enables child sockets to inherit the L3 master device index.
492 Enabling this option allows a "global" listen socket to work
493 across L3 master domains (e.g., VRFs) with connected sockets
494 derived from the listen socket to be bound to the L3 domain in
495 which the packets originated. Only valid when the kernel was
496 compiled with CONFIG_NET_L3_MASTER_DEV.
498 Default: 0 (disabled)
500 tcp_low_latency - BOOLEAN
501 This is a legacy option, it has no effect anymore.
503 tcp_max_orphans - INTEGER
504 Maximal number of TCP sockets not attached to any user file handle,
505 held by system. If this number is exceeded orphaned connections are
506 reset immediately and warning is printed. This limit exists
507 only to prevent simple DoS attacks, you _must_ not rely on this
508 or lower the limit artificially, but rather increase it
509 (probably, after increasing installed memory),
510 if network conditions require more than default value,
511 and tune network services to linger and kill such states
512 more aggressively. Let me to remind again: each orphan eats
513 up to ~64K of unswappable memory.
515 tcp_max_syn_backlog - INTEGER
516 Maximal number of remembered connection requests (SYN_RECV),
517 which have not received an acknowledgment from connecting client.
519 This is a per-listener limit.
521 The minimal value is 128 for low memory machines, and it will
522 increase in proportion to the memory of machine.
524 If server suffers from overload, try increasing this number.
526 Remember to also check /proc/sys/net/core/somaxconn
527 A SYN_RECV request socket consumes about 304 bytes of memory.
529 tcp_max_tw_buckets - INTEGER
530 Maximal number of timewait sockets held by system simultaneously.
531 If this number is exceeded time-wait socket is immediately destroyed
532 and warning is printed. This limit exists only to prevent
533 simple DoS attacks, you _must_ not lower the limit artificially,
534 but rather increase it (probably, after increasing installed memory),
535 if network conditions require more than default value.
537 tcp_mem - vector of 3 INTEGERs: min, pressure, max
538 min: below this number of pages TCP is not bothered about its
541 pressure: when amount of memory allocated by TCP exceeds this number
542 of pages, TCP moderates its memory consumption and enters memory
543 pressure mode, which is exited when memory consumption falls
546 max: number of pages allowed for queueing by all TCP sockets.
548 Defaults are calculated at boot time from amount of available
551 tcp_min_rtt_wlen - INTEGER
552 The window length of the windowed min filter to track the minimum RTT.
553 A shorter window lets a flow more quickly pick up new (higher)
554 minimum RTT when it is moved to a longer path (e.g., due to traffic
555 engineering). A longer window makes the filter more resistant to RTT
556 inflations such as transient congestion. The unit is seconds.
558 Possible values: 0 - 86400 (1 day)
562 tcp_moderate_rcvbuf - BOOLEAN
563 If set, TCP performs receive buffer auto-tuning, attempting to
564 automatically size the buffer (no greater than tcp_rmem[2]) to
565 match the size required by the path for full throughput. Enabled by
568 tcp_mtu_probing - INTEGER
569 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
573 - 1 - Disabled by default, enabled when an ICMP black hole detected
574 - 2 - Always enabled, use initial MSS of tcp_base_mss.
576 tcp_probe_interval - UNSIGNED INTEGER
577 Controls how often to start TCP Packetization-Layer Path MTU
578 Discovery reprobe. The default is reprobing every 10 minutes as
581 tcp_probe_threshold - INTEGER
582 Controls when TCP Packetization-Layer Path MTU Discovery probing
583 will stop in respect to the width of search range in bytes. Default
586 tcp_no_metrics_save - BOOLEAN
587 By default, TCP saves various connection metrics in the route cache
588 when the connection closes, so that connections established in the
589 near future can use these to set initial conditions. Usually, this
590 increases overall performance, but may sometimes cause performance
591 degradation. If set, TCP will not cache metrics on closing
594 tcp_no_ssthresh_metrics_save - BOOLEAN
595 Controls whether TCP saves ssthresh metrics in the route cache.
597 Default is 1, which disables ssthresh metrics.
599 tcp_orphan_retries - INTEGER
600 This value influences the timeout of a locally closed TCP connection,
601 when RTO retransmissions remain unacknowledged.
602 See tcp_retries2 for more details.
604 The default value is 8.
606 If your machine is a loaded WEB server,
607 you should think about lowering this value, such sockets
608 may consume significant resources. Cf. tcp_max_orphans.
610 tcp_recovery - INTEGER
611 This value is a bitmap to enable various experimental loss recovery
614 ========= =============================================================
615 RACK: 0x1 enables the RACK loss detection for fast detection of lost
616 retransmissions and tail drops. It also subsumes and disables
617 RFC6675 recovery for SACK connections.
619 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
621 RACK: 0x4 disables RACK's DUPACK threshold heuristic
622 ========= =============================================================
626 tcp_reordering - INTEGER
627 Initial reordering level of packets in a TCP stream.
628 TCP stack can then dynamically adjust flow reordering level
629 between this initial value and tcp_max_reordering
633 tcp_max_reordering - INTEGER
634 Maximal reordering level of packets in a TCP stream.
635 300 is a fairly conservative value, but you might increase it
636 if paths are using per packet load balancing (like bonding rr mode)
640 tcp_retrans_collapse - BOOLEAN
641 Bug-to-bug compatibility with some broken printers.
642 On retransmit try to send bigger packets to work around bugs in
645 tcp_retries1 - INTEGER
646 This value influences the time, after which TCP decides, that
647 something is wrong due to unacknowledged RTO retransmissions,
648 and reports this suspicion to the network layer.
649 See tcp_retries2 for more details.
651 RFC 1122 recommends at least 3 retransmissions, which is the
654 tcp_retries2 - INTEGER
655 This value influences the timeout of an alive TCP connection,
656 when RTO retransmissions remain unacknowledged.
657 Given a value of N, a hypothetical TCP connection following
658 exponential backoff with an initial RTO of TCP_RTO_MIN would
659 retransmit N times before killing the connection at the (N+1)th RTO.
661 The default value of 15 yields a hypothetical timeout of 924.6
662 seconds and is a lower bound for the effective timeout.
663 TCP will effectively time out at the first RTO which exceeds the
664 hypothetical timeout.
666 RFC 1122 recommends at least 100 seconds for the timeout,
667 which corresponds to a value of at least 8.
669 tcp_rfc1337 - BOOLEAN
670 If set, the TCP stack behaves conforming to RFC1337. If unset,
671 we are not conforming to RFC, but prevent TCP TIME_WAIT
676 tcp_rmem - vector of 3 INTEGERs: min, default, max
677 min: Minimal size of receive buffer used by TCP sockets.
678 It is guaranteed to each TCP socket, even under moderate memory
683 default: initial size of receive buffer used by TCP sockets.
684 This value overrides net.core.rmem_default used by other protocols.
685 Default: 131072 bytes.
686 This value results in initial window of 65535.
688 max: maximal size of receive buffer allowed for automatically
689 selected receiver buffers for TCP socket. This value does not override
690 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
691 automatic tuning of that socket's receive buffer size, in which
692 case this value is ignored.
693 Default: between 131072 and 6MB, depending on RAM size.
696 Enable select acknowledgments (SACKS).
698 tcp_comp_sack_delay_ns - LONG INTEGER
699 TCP tries to reduce number of SACK sent, using a timer
700 based on 5% of SRTT, capped by this sysctl, in nano seconds.
701 The default is 1ms, based on TSO autosizing period.
703 Default : 1,000,000 ns (1 ms)
705 tcp_comp_sack_slack_ns - LONG INTEGER
706 This sysctl control the slack used when arming the
707 timer used by SACK compression. This gives extra time
708 for small RTT flows, and reduces system overhead by allowing
709 opportunistic reduction of timer interrupts.
711 Default : 100,000 ns (100 us)
713 tcp_comp_sack_nr - INTEGER
714 Max number of SACK that can be compressed.
715 Using 0 disables SACK compression.
719 tcp_slow_start_after_idle - BOOLEAN
720 If set, provide RFC2861 behavior and time out the congestion
721 window after an idle period. An idle period is defined at
722 the current RTO. If unset, the congestion window will not
723 be timed out after an idle period.
728 Use the Host requirements interpretation of the TCP urgent pointer field.
729 Most hosts use the older BSD interpretation, so if you turn this on
730 Linux might not communicate correctly with them.
734 tcp_synack_retries - INTEGER
735 Number of times SYNACKs for a passive TCP connection attempt will
736 be retransmitted. Should not be higher than 255. Default value
737 is 5, which corresponds to 31seconds till the last retransmission
738 with the current initial RTO of 1second. With this the final timeout
739 for a passive TCP connection will happen after 63seconds.
741 tcp_syncookies - INTEGER
742 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
743 Send out syncookies when the syn backlog queue of a socket
744 overflows. This is to prevent against the common 'SYN flood attack'
747 Note, that syncookies is fallback facility.
748 It MUST NOT be used to help highly loaded servers to stand
749 against legal connection rate. If you see SYN flood warnings
750 in your logs, but investigation shows that they occur
751 because of overload with legal connections, you should tune
752 another parameters until this warning disappear.
753 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
755 syncookies seriously violate TCP protocol, do not allow
756 to use TCP extensions, can result in serious degradation
757 of some services (f.e. SMTP relaying), visible not by you,
758 but your clients and relays, contacting you. While you see
759 SYN flood warnings in logs not being really flooded, your server
760 is seriously misconfigured.
762 If you want to test which effects syncookies have to your
763 network connections you can set this knob to 2 to enable
764 unconditionally generation of syncookies.
766 tcp_migrate_req - BOOLEAN
767 The incoming connection is tied to a specific listening socket when
768 the initial SYN packet is received during the three-way handshake.
769 When a listener is closed, in-flight request sockets during the
770 handshake and established sockets in the accept queue are aborted.
772 If the listener has SO_REUSEPORT enabled, other listeners on the
773 same port should have been able to accept such connections. This
774 option makes it possible to migrate such child sockets to another
775 listener after close() or shutdown().
777 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
778 usually be used to define the policy to pick an alive listener.
779 Otherwise, the kernel will randomly pick an alive listener only if
780 this option is enabled.
782 Note that migration between listeners with different settings may
783 crash applications. Let's say migration happens from listener A to
784 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
785 the requests migrated from A. To avoid such a situation, cancel
786 migration by returning SK_DROP in the type of eBPF program, or
791 tcp_fastopen - INTEGER
792 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
795 The client support is enabled by flag 0x1 (on by default). The client
796 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
797 rather than connect() to send data in SYN.
799 The server support is enabled by flag 0x2 (off by default). Then
800 either enable for all listeners with another flag (0x400) or
801 enable individual listeners via TCP_FASTOPEN socket option with
802 the option value being the length of the syn-data backlog.
804 The values (bitmap) are
806 ===== ======== ======================================================
807 0x1 (client) enables sending data in the opening SYN on the client.
808 0x2 (server) enables the server support, i.e., allowing data in
809 a SYN packet to be accepted and passed to the
810 application before 3-way handshake finishes.
811 0x4 (client) send data in the opening SYN regardless of cookie
812 availability and without a cookie option.
813 0x200 (server) accept data-in-SYN w/o any cookie option present.
814 0x400 (server) enable all listeners to support Fast Open by
815 default without explicit TCP_FASTOPEN socket option.
816 ===== ======== ======================================================
820 Note that additional client or server features are only
821 effective if the basic support (0x1 and 0x2) are enabled respectively.
823 tcp_fastopen_blackhole_timeout_sec - INTEGER
824 Initial time period in second to disable Fastopen on active TCP sockets
825 when a TFO firewall blackhole issue happens.
826 This time period will grow exponentially when more blackhole issues
827 get detected right after Fastopen is re-enabled and will reset to
828 initial value when the blackhole issue goes away.
829 0 to disable the blackhole detection.
831 By default, it is set to 0 (feature is disabled).
833 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
834 The list consists of a primary key and an optional backup key. The
835 primary key is used for both creating and validating cookies, while the
836 optional backup key is only used for validating cookies. The purpose of
837 the backup key is to maximize TFO validation when keys are rotated.
839 A randomly chosen primary key may be configured by the kernel if
840 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
841 TCP_FASTOPEN setsockopt() optname is set and a key has not been
842 previously configured via sysctl. If keys are configured via
843 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
844 per-socket keys will be used instead of any keys that are specified via
847 A key is specified as 4 8-digit hexadecimal integers which are separated
848 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
849 omitted. A primary and a backup key may be specified by separating them
850 by a comma. If only one key is specified, it becomes the primary key and
851 any previously configured backup keys are removed.
853 tcp_syn_retries - INTEGER
854 Number of times initial SYNs for an active TCP connection attempt
855 will be retransmitted. Should not be higher than 127. Default value
856 is 6, which corresponds to 63seconds till the last retransmission
857 with the current initial RTO of 1second. With this the final timeout
858 for an active TCP connection attempt will happen after 127seconds.
860 tcp_timestamps - INTEGER
861 Enable timestamps as defined in RFC1323.
864 - 1: Enable timestamps as defined in RFC1323 and use random offset for
865 each connection rather than only using the current time.
866 - 2: Like 1, but without random offsets.
870 tcp_min_tso_segs - INTEGER
871 Minimal number of segments per TSO frame.
873 Since linux-3.12, TCP does an automatic sizing of TSO frames,
874 depending on flow rate, instead of filling 64Kbytes packets.
875 For specific usages, it's possible to force TCP to build big
876 TSO frames. Note that TCP stack might split too big TSO packets
877 if available window is too small.
881 tcp_pacing_ss_ratio - INTEGER
882 sk->sk_pacing_rate is set by TCP stack using a ratio applied
883 to current rate. (current_rate = cwnd * mss / srtt)
884 If TCP is in slow start, tcp_pacing_ss_ratio is applied
885 to let TCP probe for bigger speeds, assuming cwnd can be
886 doubled every other RTT.
890 tcp_pacing_ca_ratio - INTEGER
891 sk->sk_pacing_rate is set by TCP stack using a ratio applied
892 to current rate. (current_rate = cwnd * mss / srtt)
893 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
894 is applied to conservatively probe for bigger throughput.
898 tcp_tso_win_divisor - INTEGER
899 This allows control over what percentage of the congestion window
900 can be consumed by a single TSO frame.
901 The setting of this parameter is a choice between burstiness and
902 building larger TSO frames.
906 tcp_tw_reuse - INTEGER
907 Enable reuse of TIME-WAIT sockets for new connections when it is
908 safe from protocol viewpoint.
912 - 2 - enable for loopback traffic only
914 It should not be changed without advice/request of technical
919 tcp_window_scaling - BOOLEAN
920 Enable window scaling as defined in RFC1323.
922 tcp_wmem - vector of 3 INTEGERs: min, default, max
923 min: Amount of memory reserved for send buffers for TCP sockets.
924 Each TCP socket has rights to use it due to fact of its birth.
928 default: initial size of send buffer used by TCP sockets. This
929 value overrides net.core.wmem_default used by other protocols.
931 It is usually lower than net.core.wmem_default.
935 max: Maximal amount of memory allowed for automatically tuned
936 send buffers for TCP sockets. This value does not override
937 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
938 automatic tuning of that socket's send buffer size, in which case
939 this value is ignored.
941 Default: between 64K and 4MB, depending on RAM size.
943 tcp_notsent_lowat - UNSIGNED INTEGER
944 A TCP socket can control the amount of unsent bytes in its write queue,
945 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
946 reports POLLOUT events if the amount of unsent bytes is below a per
947 socket value, and if the write queue is not full. sendmsg() will
948 also not add new buffers if the limit is hit.
950 This global variable controls the amount of unsent data for
951 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
952 to the global variable has immediate effect.
954 Default: UINT_MAX (0xFFFFFFFF)
956 tcp_workaround_signed_windows - BOOLEAN
957 If set, assume no receipt of a window scaling option means the
958 remote TCP is broken and treats the window as a signed quantity.
959 If unset, assume the remote TCP is not broken even if we do
960 not receive a window scaling option from them.
964 tcp_thin_linear_timeouts - BOOLEAN
965 Enable dynamic triggering of linear timeouts for thin streams.
966 If set, a check is performed upon retransmission by timeout to
967 determine if the stream is thin (less than 4 packets in flight).
968 As long as the stream is found to be thin, up to 6 linear
969 timeouts may be performed before exponential backoff mode is
970 initiated. This improves retransmission latency for
971 non-aggressive thin streams, often found to be time-dependent.
972 For more information on thin streams, see
973 Documentation/networking/tcp-thin.rst
977 tcp_limit_output_bytes - INTEGER
978 Controls TCP Small Queue limit per tcp socket.
979 TCP bulk sender tends to increase packets in flight until it
980 gets losses notifications. With SNDBUF autotuning, this can
981 result in a large amount of packets queued on the local machine
982 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
983 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
984 limits the number of bytes on qdisc or device to reduce artificial
985 RTT/cwnd and reduce bufferbloat.
987 Default: 1048576 (16 * 65536)
989 tcp_challenge_ack_limit - INTEGER
990 Limits number of Challenge ACK sent per second, as recommended
991 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
994 tcp_rx_skb_cache - BOOLEAN
995 Controls a per TCP socket cache of one skb, that might help
996 performance of some workloads. This might be dangerous
997 on systems with a lot of TCP sockets, since it increases
1000 Default: 0 (disabled)
1005 udp_l3mdev_accept - BOOLEAN
1006 Enabling this option allows a "global" bound socket to work
1007 across L3 master domains (e.g., VRFs) with packets capable of
1008 being received regardless of the L3 domain in which they
1009 originated. Only valid when the kernel was compiled with
1010 CONFIG_NET_L3_MASTER_DEV.
1012 Default: 0 (disabled)
1014 udp_mem - vector of 3 INTEGERs: min, pressure, max
1015 Number of pages allowed for queueing by all UDP sockets.
1017 min: Below this number of pages UDP is not bothered about its
1018 memory appetite. When amount of memory allocated by UDP exceeds
1019 this number, UDP starts to moderate memory usage.
1021 pressure: This value was introduced to follow format of tcp_mem.
1023 max: Number of pages allowed for queueing by all UDP sockets.
1025 Default is calculated at boot time from amount of available memory.
1027 udp_rmem_min - INTEGER
1028 Minimal size of receive buffer used by UDP sockets in moderation.
1029 Each UDP socket is able to use the size for receiving data, even if
1030 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1034 udp_wmem_min - INTEGER
1035 Minimal size of send buffer used by UDP sockets in moderation.
1036 Each UDP socket is able to use the size for sending data, even if
1037 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1044 raw_l3mdev_accept - BOOLEAN
1045 Enabling this option allows a "global" bound socket to work
1046 across L3 master domains (e.g., VRFs) with packets capable of
1047 being received regardless of the L3 domain in which they
1048 originated. Only valid when the kernel was compiled with
1049 CONFIG_NET_L3_MASTER_DEV.
1051 Default: 1 (enabled)
1056 cipso_cache_enable - BOOLEAN
1057 If set, enable additions to and lookups from the CIPSO label mapping
1058 cache. If unset, additions are ignored and lookups always result in a
1059 miss. However, regardless of the setting the cache is still
1060 invalidated when required when means you can safely toggle this on and
1061 off and the cache will always be "safe".
1065 cipso_cache_bucket_size - INTEGER
1066 The CIPSO label cache consists of a fixed size hash table with each
1067 hash bucket containing a number of cache entries. This variable limits
1068 the number of entries in each hash bucket; the larger the value is, the
1069 more CIPSO label mappings that can be cached. When the number of
1070 entries in a given hash bucket reaches this limit adding new entries
1071 causes the oldest entry in the bucket to be removed to make room.
1075 cipso_rbm_optfmt - BOOLEAN
1076 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1077 the CIPSO draft specification (see Documentation/netlabel for details).
1078 This means that when set the CIPSO tag will be padded with empty
1079 categories in order to make the packet data 32-bit aligned.
1083 cipso_rbm_structvalid - BOOLEAN
1084 If set, do a very strict check of the CIPSO option when
1085 ip_options_compile() is called. If unset, relax the checks done during
1086 ip_options_compile(). Either way is "safe" as errors are caught else
1087 where in the CIPSO processing code but setting this to 0 (False) should
1088 result in less work (i.e. it should be faster) but could cause problems
1089 with other implementations that require strict checking.
1096 ip_local_port_range - 2 INTEGERS
1097 Defines the local port range that is used by TCP and UDP to
1098 choose the local port. The first number is the first, the
1099 second the last local port number.
1100 If possible, it is better these numbers have different parity
1101 (one even and one odd value).
1102 Must be greater than or equal to ip_unprivileged_port_start.
1103 The default values are 32768 and 60999 respectively.
1105 ip_local_reserved_ports - list of comma separated ranges
1106 Specify the ports which are reserved for known third-party
1107 applications. These ports will not be used by automatic port
1108 assignments (e.g. when calling connect() or bind() with port
1109 number 0). Explicit port allocation behavior is unchanged.
1111 The format used for both input and output is a comma separated
1112 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1113 10). Writing to the file will clear all previously reserved
1114 ports and update the current list with the one given in the
1117 Note that ip_local_port_range and ip_local_reserved_ports
1118 settings are independent and both are considered by the kernel
1119 when determining which ports are available for automatic port
1122 You can reserve ports which are not in the current
1123 ip_local_port_range, e.g.::
1125 $ cat /proc/sys/net/ipv4/ip_local_port_range
1127 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1130 although this is redundant. However such a setting is useful
1131 if later the port range is changed to a value that will
1132 include the reserved ports. Also keep in mind, that overlapping
1133 of these ranges may affect probability of selecting ephemeral
1134 ports which are right after block of reserved ports.
1138 ip_unprivileged_port_start - INTEGER
1139 This is a per-namespace sysctl. It defines the first
1140 unprivileged port in the network namespace. Privileged ports
1141 require root or CAP_NET_BIND_SERVICE in order to bind to them.
1142 To disable all privileged ports, set this to 0. They must not
1143 overlap with the ip_local_port_range.
1147 ip_nonlocal_bind - BOOLEAN
1148 If set, allows processes to bind() to non-local IP addresses,
1149 which can be quite useful - but may break some applications.
1153 ip_autobind_reuse - BOOLEAN
1154 By default, bind() does not select the ports automatically even if
1155 the new socket and all sockets bound to the port have SO_REUSEADDR.
1156 ip_autobind_reuse allows bind() to reuse the port and this is useful
1157 when you use bind()+connect(), but may break some applications.
1158 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1159 option should only be set by experts.
1162 ip_dynaddr - INTEGER
1163 If set non-zero, enables support for dynamic addresses.
1164 If set to a non-zero value larger than 1, a kernel log
1165 message will be printed when dynamic address rewriting
1170 ip_early_demux - BOOLEAN
1171 Optimize input packet processing down to one demux for
1172 certain kinds of local sockets. Currently we only do this
1173 for established TCP and connected UDP sockets.
1175 It may add an additional cost for pure routing workloads that
1176 reduces overall throughput, in such case you should disable it.
1180 ping_group_range - 2 INTEGERS
1181 Restrict ICMP_PROTO datagram sockets to users in the group range.
1182 The default is "1 0", meaning, that nobody (not even root) may
1183 create ping sockets. Setting it to "100 100" would grant permissions
1184 to the single group. "0 4294967295" would enable it for the world, "100
1185 4294967295" would enable it for the users, but not daemons.
1187 tcp_early_demux - BOOLEAN
1188 Enable early demux for established TCP sockets.
1192 udp_early_demux - BOOLEAN
1193 Enable early demux for connected UDP sockets. Disable this if
1194 your system could experience more unconnected load.
1198 icmp_echo_ignore_all - BOOLEAN
1199 If set non-zero, then the kernel will ignore all ICMP ECHO
1200 requests sent to it.
1204 icmp_echo_enable_probe - BOOLEAN
1205 If set to one, then the kernel will respond to RFC 8335 PROBE
1206 requests sent to it.
1210 icmp_echo_ignore_broadcasts - BOOLEAN
1211 If set non-zero, then the kernel will ignore all ICMP ECHO and
1212 TIMESTAMP requests sent to it via broadcast/multicast.
1216 icmp_ratelimit - INTEGER
1217 Limit the maximal rates for sending ICMP packets whose type matches
1218 icmp_ratemask (see below) to specific targets.
1219 0 to disable any limiting,
1220 otherwise the minimal space between responses in milliseconds.
1221 Note that another sysctl, icmp_msgs_per_sec limits the number
1222 of ICMP packets sent on all targets.
1226 icmp_msgs_per_sec - INTEGER
1227 Limit maximal number of ICMP packets sent per second from this host.
1228 Only messages whose type matches icmp_ratemask (see below) are
1229 controlled by this limit. For security reasons, the precise count
1230 of messages per second is randomized.
1234 icmp_msgs_burst - INTEGER
1235 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1236 while icmp_msgs_burst controls the burst size of these packets.
1237 For security reasons, the precise burst size is randomized.
1241 icmp_ratemask - INTEGER
1242 Mask made of ICMP types for which rates are being limited.
1244 Significant bits: IHGFEDCBA9876543210
1246 Default mask: 0000001100000011000 (6168)
1248 Bit definitions (see include/linux/icmp.h):
1250 = =========================
1252 3 Destination Unreachable [1]_
1253 4 Source Quench [1]_
1256 B Time Exceeded [1]_
1257 C Parameter Problem [1]_
1262 H Address Mask Request
1263 I Address Mask Reply
1264 = =========================
1266 .. [1] These are rate limited by default (see default mask above)
1268 icmp_ignore_bogus_error_responses - BOOLEAN
1269 Some routers violate RFC1122 by sending bogus responses to broadcast
1270 frames. Such violations are normally logged via a kernel warning.
1271 If this is set to TRUE, the kernel will not give such warnings, which
1272 will avoid log file clutter.
1276 icmp_errors_use_inbound_ifaddr - BOOLEAN
1278 If zero, icmp error messages are sent with the primary address of
1279 the exiting interface.
1281 If non-zero, the message will be sent with the primary address of
1282 the interface that received the packet that caused the icmp error.
1283 This is the behaviour many network administrators will expect from
1284 a router. And it can make debugging complicated network layouts
1287 Note that if no primary address exists for the interface selected,
1288 then the primary address of the first non-loopback interface that
1289 has one will be used regardless of this setting.
1293 igmp_max_memberships - INTEGER
1294 Change the maximum number of multicast groups we can subscribe to.
1297 Theoretical maximum value is bounded by having to send a membership
1298 report in a single datagram (i.e. the report can't span multiple
1299 datagrams, or risk confusing the switch and leaving groups you don't
1302 The number of supported groups 'M' is bounded by the number of group
1303 report entries you can fit into a single datagram of 65535 bytes.
1305 M = 65536-sizeof (ip header)/(sizeof(Group record))
1307 Group records are variable length, with a minimum of 12 bytes.
1308 So net.ipv4.igmp_max_memberships should not be set higher than:
1310 (65536-24) / 12 = 5459
1312 The value 5459 assumes no IP header options, so in practice
1313 this number may be lower.
1315 igmp_max_msf - INTEGER
1316 Maximum number of addresses allowed in the source filter list for a
1322 Controls the IGMP query robustness variable (see RFC2236 8.1).
1324 Default: 2 (as specified by RFC2236 8.1)
1326 Minimum: 1 (as specified by RFC6636 4.5)
1328 force_igmp_version - INTEGER
1329 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1330 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1331 Present timer expires.
1332 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1333 receive IGMPv2/v3 query.
1334 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1335 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1336 - 3 - Enforce to use IGMP version 3. The same react with default 0.
1340 this is not the same with force_mld_version because IGMPv3 RFC3376
1341 Security Considerations does not have clear description that we could
1342 ignore other version messages completely as MLDv2 RFC3810. So make
1343 this value as default 0 is recommended.
1345 ``conf/interface/*``
1346 changes special settings per interface (where
1347 interface" is the name of your network interface)
1350 is special, changes the settings for all interfaces
1352 log_martians - BOOLEAN
1353 Log packets with impossible addresses to kernel log.
1354 log_martians for the interface will be enabled if at least one of
1355 conf/{all,interface}/log_martians is set to TRUE,
1356 it will be disabled otherwise
1358 accept_redirects - BOOLEAN
1359 Accept ICMP redirect messages.
1360 accept_redirects for the interface will be enabled if:
1362 - both conf/{all,interface}/accept_redirects are TRUE in the case
1363 forwarding for the interface is enabled
1367 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1368 case forwarding for the interface is disabled
1370 accept_redirects for the interface will be disabled otherwise
1377 forwarding - BOOLEAN
1378 Enable IP forwarding on this interface. This controls whether packets
1379 received _on_ this interface can be forwarded.
1381 mc_forwarding - BOOLEAN
1382 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1383 and a multicast routing daemon is required.
1384 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1385 routing for the interface
1388 Integer value used to differentiate the devices by the medium they
1389 are attached to. Two devices can have different id values when
1390 the broadcast packets are received only on one of them.
1391 The default value 0 means that the device is the only interface
1392 to its medium, value of -1 means that medium is not known.
1394 Currently, it is used to change the proxy_arp behavior:
1395 the proxy_arp feature is enabled for packets forwarded between
1396 two devices attached to different media.
1401 proxy_arp for the interface will be enabled if at least one of
1402 conf/{all,interface}/proxy_arp is set to TRUE,
1403 it will be disabled otherwise
1405 proxy_arp_pvlan - BOOLEAN
1406 Private VLAN proxy arp.
1408 Basically allow proxy arp replies back to the same interface
1409 (from which the ARP request/solicitation was received).
1411 This is done to support (ethernet) switch features, like RFC
1412 3069, where the individual ports are NOT allowed to
1413 communicate with each other, but they are allowed to talk to
1414 the upstream router. As described in RFC 3069, it is possible
1415 to allow these hosts to communicate through the upstream
1416 router by proxy_arp'ing. Don't need to be used together with
1419 This technology is known by different names:
1421 In RFC 3069 it is called VLAN Aggregation.
1422 Cisco and Allied Telesyn call it Private VLAN.
1423 Hewlett-Packard call it Source-Port filtering or port-isolation.
1424 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1426 shared_media - BOOLEAN
1427 Send(router) or accept(host) RFC1620 shared media redirects.
1428 Overrides secure_redirects.
1430 shared_media for the interface will be enabled if at least one of
1431 conf/{all,interface}/shared_media is set to TRUE,
1432 it will be disabled otherwise
1436 secure_redirects - BOOLEAN
1437 Accept ICMP redirect messages only to gateways listed in the
1438 interface's current gateway list. Even if disabled, RFC1122 redirect
1441 Overridden by shared_media.
1443 secure_redirects for the interface will be enabled if at least one of
1444 conf/{all,interface}/secure_redirects is set to TRUE,
1445 it will be disabled otherwise
1449 send_redirects - BOOLEAN
1450 Send redirects, if router.
1452 send_redirects for the interface will be enabled if at least one of
1453 conf/{all,interface}/send_redirects is set to TRUE,
1454 it will be disabled otherwise
1458 bootp_relay - BOOLEAN
1459 Accept packets with source address 0.b.c.d destined
1460 not to this host as local ones. It is supposed, that
1461 BOOTP relay daemon will catch and forward such packets.
1462 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1467 Not Implemented Yet.
1469 accept_source_route - BOOLEAN
1470 Accept packets with SRR option.
1471 conf/all/accept_source_route must also be set to TRUE to accept packets
1472 with SRR option on the interface
1479 accept_local - BOOLEAN
1480 Accept packets with local source addresses. In combination with
1481 suitable routing, this can be used to direct packets between two
1482 local interfaces over the wire and have them accepted properly.
1485 route_localnet - BOOLEAN
1486 Do not consider loopback addresses as martian source or destination
1487 while routing. This enables the use of 127/8 for local routing purposes.
1492 - 0 - No source validation.
1493 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1494 Each incoming packet is tested against the FIB and if the interface
1495 is not the best reverse path the packet check will fail.
1496 By default failed packets are discarded.
1497 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1498 Each incoming packet's source address is also tested against the FIB
1499 and if the source address is not reachable via any interface
1500 the packet check will fail.
1502 Current recommended practice in RFC3704 is to enable strict mode
1503 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1504 or other complicated routing, then loose mode is recommended.
1506 The max value from conf/{all,interface}/rp_filter is used
1507 when doing source validation on the {interface}.
1509 Default value is 0. Note that some distributions enable it
1512 src_valid_mark - BOOLEAN
1513 - 0 - The fwmark of the packet is not included in reverse path
1514 route lookup. This allows for asymmetric routing configurations
1515 utilizing the fwmark in only one direction, e.g., transparent
1518 - 1 - The fwmark of the packet is included in reverse path route
1519 lookup. This permits rp_filter to function when the fwmark is
1520 used for routing traffic in both directions.
1522 This setting also affects the utilization of fmwark when
1523 performing source address selection for ICMP replies, or
1524 determining addresses stored for the IPOPT_TS_TSANDADDR and
1525 IPOPT_RR IP options.
1527 The max value from conf/{all,interface}/src_valid_mark is used.
1531 arp_filter - BOOLEAN
1532 - 1 - Allows you to have multiple network interfaces on the same
1533 subnet, and have the ARPs for each interface be answered
1534 based on whether or not the kernel would route a packet from
1535 the ARP'd IP out that interface (therefore you must use source
1536 based routing for this to work). In other words it allows control
1537 of which cards (usually 1) will respond to an arp request.
1539 - 0 - (default) The kernel can respond to arp requests with addresses
1540 from other interfaces. This may seem wrong but it usually makes
1541 sense, because it increases the chance of successful communication.
1542 IP addresses are owned by the complete host on Linux, not by
1543 particular interfaces. Only for more complex setups like load-
1544 balancing, does this behaviour cause problems.
1546 arp_filter for the interface will be enabled if at least one of
1547 conf/{all,interface}/arp_filter is set to TRUE,
1548 it will be disabled otherwise
1550 arp_announce - INTEGER
1551 Define different restriction levels for announcing the local
1552 source IP address from IP packets in ARP requests sent on
1555 - 0 - (default) Use any local address, configured on any interface
1556 - 1 - Try to avoid local addresses that are not in the target's
1557 subnet for this interface. This mode is useful when target
1558 hosts reachable via this interface require the source IP
1559 address in ARP requests to be part of their logical network
1560 configured on the receiving interface. When we generate the
1561 request we will check all our subnets that include the
1562 target IP and will preserve the source address if it is from
1563 such subnet. If there is no such subnet we select source
1564 address according to the rules for level 2.
1565 - 2 - Always use the best local address for this target.
1566 In this mode we ignore the source address in the IP packet
1567 and try to select local address that we prefer for talks with
1568 the target host. Such local address is selected by looking
1569 for primary IP addresses on all our subnets on the outgoing
1570 interface that include the target IP address. If no suitable
1571 local address is found we select the first local address
1572 we have on the outgoing interface or on all other interfaces,
1573 with the hope we will receive reply for our request and
1574 even sometimes no matter the source IP address we announce.
1576 The max value from conf/{all,interface}/arp_announce is used.
1578 Increasing the restriction level gives more chance for
1579 receiving answer from the resolved target while decreasing
1580 the level announces more valid sender's information.
1582 arp_ignore - INTEGER
1583 Define different modes for sending replies in response to
1584 received ARP requests that resolve local target IP addresses:
1586 - 0 - (default): reply for any local target IP address, configured
1588 - 1 - reply only if the target IP address is local address
1589 configured on the incoming interface
1590 - 2 - reply only if the target IP address is local address
1591 configured on the incoming interface and both with the
1592 sender's IP address are part from same subnet on this interface
1593 - 3 - do not reply for local addresses configured with scope host,
1594 only resolutions for global and link addresses are replied
1596 - 8 - do not reply for all local addresses
1598 The max value from conf/{all,interface}/arp_ignore is used
1599 when ARP request is received on the {interface}
1601 arp_notify - BOOLEAN
1602 Define mode for notification of address and device changes.
1604 == ==========================================================
1605 0 (default): do nothing
1606 1 Generate gratuitous arp requests when device is brought up
1607 or hardware address changes.
1608 == ==========================================================
1610 arp_accept - BOOLEAN
1611 Define behavior for gratuitous ARP frames who's IP is not
1612 already present in the ARP table:
1614 - 0 - don't create new entries in the ARP table
1615 - 1 - create new entries in the ARP table
1617 Both replies and requests type gratuitous arp will trigger the
1618 ARP table to be updated, if this setting is on.
1620 If the ARP table already contains the IP address of the
1621 gratuitous arp frame, the arp table will be updated regardless
1622 if this setting is on or off.
1624 mcast_solicit - INTEGER
1625 The maximum number of multicast probes in INCOMPLETE state,
1626 when the associated hardware address is unknown. Defaults
1629 ucast_solicit - INTEGER
1630 The maximum number of unicast probes in PROBE state, when
1631 the hardware address is being reconfirmed. Defaults to 3.
1633 app_solicit - INTEGER
1634 The maximum number of probes to send to the user space ARP daemon
1635 via netlink before dropping back to multicast probes (see
1636 mcast_resolicit). Defaults to 0.
1638 mcast_resolicit - INTEGER
1639 The maximum number of multicast probes after unicast and
1640 app probes in PROBE state. Defaults to 0.
1642 disable_policy - BOOLEAN
1643 Disable IPSEC policy (SPD) for this interface
1645 disable_xfrm - BOOLEAN
1646 Disable IPSEC encryption on this interface, whatever the policy
1648 igmpv2_unsolicited_report_interval - INTEGER
1649 The interval in milliseconds in which the next unsolicited
1650 IGMPv1 or IGMPv2 report retransmit will take place.
1652 Default: 10000 (10 seconds)
1654 igmpv3_unsolicited_report_interval - INTEGER
1655 The interval in milliseconds in which the next unsolicited
1656 IGMPv3 report retransmit will take place.
1658 Default: 1000 (1 seconds)
1660 ignore_routes_with_linkdown - BOOLEAN
1661 Ignore routes whose link is down when performing a FIB lookup.
1663 promote_secondaries - BOOLEAN
1664 When a primary IP address is removed from this interface
1665 promote a corresponding secondary IP address instead of
1666 removing all the corresponding secondary IP addresses.
1668 drop_unicast_in_l2_multicast - BOOLEAN
1669 Drop any unicast IP packets that are received in link-layer
1670 multicast (or broadcast) frames.
1672 This behavior (for multicast) is actually a SHOULD in RFC
1673 1122, but is disabled by default for compatibility reasons.
1677 drop_gratuitous_arp - BOOLEAN
1678 Drop all gratuitous ARP frames, for example if there's a known
1679 good ARP proxy on the network and such frames need not be used
1680 (or in the case of 802.11, must not be used to prevent attacks.)
1686 Allows you to write a number, which can be used as required.
1690 xfrm4_gc_thresh - INTEGER
1691 (Obsolete since linux-4.14)
1692 The threshold at which we will start garbage collecting for IPv4
1693 destination cache entries. At twice this value the system will
1694 refuse new allocations.
1696 igmp_link_local_mcast_reports - BOOLEAN
1697 Enable IGMP reports for link local multicast groups in the
1703 kuznet@ms2.inr.ac.ru
1710 delon.nicolas@wanadoo.fr
1715 /proc/sys/net/ipv6/* Variables
1716 ==============================
1718 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1719 apply to IPv6 [XXX?].
1721 bindv6only - BOOLEAN
1722 Default value for IPV6_V6ONLY socket option,
1723 which restricts use of the IPv6 socket to IPv6 communication
1726 - TRUE: disable IPv4-mapped address feature
1727 - FALSE: enable IPv4-mapped address feature
1729 Default: FALSE (as specified in RFC3493)
1731 flowlabel_consistency - BOOLEAN
1732 Protect the consistency (and unicity) of flow label.
1733 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1741 auto_flowlabels - INTEGER
1742 Automatically generate flow labels based on a flow hash of the
1743 packet. This allows intermediate devices, such as routers, to
1744 identify packet flows for mechanisms like Equal Cost Multipath
1745 Routing (see RFC 6438).
1747 = ===========================================================
1748 0 automatic flow labels are completely disabled
1749 1 automatic flow labels are enabled by default, they can be
1750 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1752 2 automatic flow labels are allowed, they may be enabled on a
1753 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1754 3 automatic flow labels are enabled and enforced, they cannot
1755 be disabled by the socket option
1756 = ===========================================================
1760 flowlabel_state_ranges - BOOLEAN
1761 Split the flow label number space into two ranges. 0-0x7FFFF is
1762 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1763 is reserved for stateless flow labels as described in RFC6437.
1770 flowlabel_reflect - INTEGER
1771 Control flow label reflection. Needed for Path MTU
1772 Discovery to work with Equal Cost Multipath Routing in anycast
1773 environments. See RFC 7690 and:
1774 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1778 - 1: enabled for established flows
1780 Note that this prevents automatic flowlabel changes, as done
1781 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
1782 and "tcp: Change txhash on every SYN and RTO retransmit"
1784 - 2: enabled for TCP RESET packets (no active listener)
1785 If set, a RST packet sent in response to a SYN packet on a closed
1786 port will reflect the incoming flow label.
1788 - 4: enabled for ICMPv6 echo reply messages.
1792 fib_multipath_hash_policy - INTEGER
1793 Controls which hash policy to use for multipath routes.
1795 Default: 0 (Layer 3)
1799 - 0 - Layer 3 (source and destination addresses plus flow label)
1800 - 1 - Layer 4 (standard 5-tuple)
1801 - 2 - Layer 3 or inner Layer 3 if present
1802 - 3 - Custom multipath hash. Fields used for multipath hash calculation
1803 are determined by fib_multipath_hash_fields sysctl
1805 fib_multipath_hash_fields - UNSIGNED INTEGER
1806 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
1807 fields used for multipath hash calculation are determined by this
1810 This value is a bitmask which enables various fields for multipath hash
1813 Possible fields are:
1815 ====== ============================
1816 0x0001 Source IP address
1817 0x0002 Destination IP address
1821 0x0020 Destination port
1822 0x0040 Inner source IP address
1823 0x0080 Inner destination IP address
1824 0x0100 Inner IP protocol
1825 0x0200 Inner Flow Label
1826 0x0400 Inner source port
1827 0x0800 Inner destination port
1828 ====== ============================
1830 Default: 0x0007 (source IP, destination IP and IP protocol)
1832 anycast_src_echo_reply - BOOLEAN
1833 Controls the use of anycast addresses as source addresses for ICMPv6
1841 idgen_delay - INTEGER
1842 Controls the delay in seconds after which time to retry
1843 privacy stable address generation if a DAD conflict is
1846 Default: 1 (as specified in RFC7217)
1848 idgen_retries - INTEGER
1849 Controls the number of retries to generate a stable privacy
1850 address if a DAD conflict is detected.
1852 Default: 3 (as specified in RFC7217)
1855 Controls the MLD query robustness variable (see RFC3810 9.1).
1857 Default: 2 (as specified by RFC3810 9.1)
1859 Minimum: 1 (as specified by RFC6636 4.5)
1861 max_dst_opts_number - INTEGER
1862 Maximum number of non-padding TLVs allowed in a Destination
1863 options extension header. If this value is less than zero
1864 then unknown options are disallowed and the number of known
1865 TLVs allowed is the absolute value of this number.
1869 max_hbh_opts_number - INTEGER
1870 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1871 options extension header. If this value is less than zero
1872 then unknown options are disallowed and the number of known
1873 TLVs allowed is the absolute value of this number.
1877 max_dst_opts_length - INTEGER
1878 Maximum length allowed for a Destination options extension
1881 Default: INT_MAX (unlimited)
1883 max_hbh_length - INTEGER
1884 Maximum length allowed for a Hop-by-Hop options extension
1887 Default: INT_MAX (unlimited)
1889 skip_notify_on_dev_down - BOOLEAN
1890 Controls whether an RTM_DELROUTE message is generated for routes
1891 removed when a device is taken down or deleted. IPv4 does not
1892 generate this message; IPv6 does by default. Setting this sysctl
1893 to true skips the message, making IPv4 and IPv6 on par in relying
1894 on userspace caches to track link events and evict routes.
1896 Default: false (generate message)
1898 nexthop_compat_mode - BOOLEAN
1899 New nexthop API provides a means for managing nexthops independent of
1900 prefixes. Backwards compatibilty with old route format is enabled by
1901 default which means route dumps and notifications contain the new
1902 nexthop attribute but also the full, expanded nexthop definition.
1903 Further, updates or deletes of a nexthop configuration generate route
1904 notifications for each fib entry using the nexthop. Once a system
1905 understands the new API, this sysctl can be disabled to achieve full
1906 performance benefits of the new API by disabling the nexthop expansion
1907 and extraneous notifications.
1908 Default: true (backward compat mode)
1910 fib_notify_on_flag_change - INTEGER
1911 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
1912 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
1914 After installing a route to the kernel, user space receives an
1915 acknowledgment, which means the route was installed in the kernel,
1916 but not necessarily in hardware.
1917 It is also possible for a route already installed in hardware to change
1918 its action and therefore its flags. For example, a host route that is
1919 trapping packets can be "promoted" to perform decapsulation following
1920 the installation of an IPinIP/VXLAN tunnel.
1921 The notifications will indicate to user-space the state of the route.
1923 Default: 0 (Do not emit notifications.)
1927 - 0 - Do not emit notifications.
1928 - 1 - Emit notifications.
1929 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
1932 Define the IOAM id of this node. Uses only 24 bits out of 32 in total.
1939 ioam6_id_wide - LONG INTEGER
1940 Define the wide IOAM id of this node. Uses only 56 bits out of 64 in
1941 total. Can be different from ioam6_id.
1944 Max: 0xFFFFFFFFFFFFFF
1946 Default: 0xFFFFFFFFFFFFFF
1950 ip6frag_high_thresh - INTEGER
1951 Maximum memory used to reassemble IPv6 fragments. When
1952 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1953 the fragment handler will toss packets until ip6frag_low_thresh
1956 ip6frag_low_thresh - INTEGER
1957 See ip6frag_high_thresh
1959 ip6frag_time - INTEGER
1960 Time in seconds to keep an IPv6 fragment in memory.
1963 Change the interface-specific default settings.
1965 These settings would be used during creating new interfaces.
1969 Change all the interface-specific settings.
1971 [XXX: Other special features than forwarding?]
1973 conf/all/disable_ipv6 - BOOLEAN
1974 Changing this value is same as changing ``conf/default/disable_ipv6``
1975 setting and also all per-interface ``disable_ipv6`` settings to the same
1978 Reading this value does not have any particular meaning. It does not say
1979 whether IPv6 support is enabled or disabled. Returned value can be 1
1980 also in the case when some interface has ``disable_ipv6`` set to 0 and
1981 has configured IPv6 addresses.
1983 conf/all/forwarding - BOOLEAN
1984 Enable global IPv6 forwarding between all interfaces.
1986 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1987 to control which interfaces may forward packets and which not.
1989 This also sets all interfaces' Host/Router setting
1990 'forwarding' to the specified value. See below for details.
1992 This referred to as global forwarding.
1997 fwmark_reflect - BOOLEAN
1998 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1999 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
2000 If unset, these packets have a fwmark of zero. If set, they have the
2001 fwmark of the packet they are replying to.
2005 ``conf/interface/*``:
2006 Change special settings per interface.
2008 The functional behaviour for certain settings is different
2009 depending on whether local forwarding is enabled or not.
2012 Accept Router Advertisements; autoconfigure using them.
2014 It also determines whether or not to transmit Router
2015 Solicitations. If and only if the functional setting is to
2016 accept Router Advertisements, Router Solicitations will be
2019 Possible values are:
2021 == ===========================================================
2022 0 Do not accept Router Advertisements.
2023 1 Accept Router Advertisements if forwarding is disabled.
2024 2 Overrule forwarding behaviour. Accept Router Advertisements
2025 even if forwarding is enabled.
2026 == ===========================================================
2030 - enabled if local forwarding is disabled.
2031 - disabled if local forwarding is enabled.
2033 accept_ra_defrtr - BOOLEAN
2034 Learn default router in Router Advertisement.
2038 - enabled if accept_ra is enabled.
2039 - disabled if accept_ra is disabled.
2041 ra_defrtr_metric - UNSIGNED INTEGER
2042 Route metric for default route learned in Router Advertisement. This value
2043 will be assigned as metric for the default route learned via IPv6 Router
2044 Advertisement. Takes affect only if accept_ra_defrtr is enabled.
2049 Default: IP6_RT_PRIO_USER i.e. 1024.
2051 accept_ra_from_local - BOOLEAN
2052 Accept RA with source-address that is found on local machine
2053 if the RA is otherwise proper and able to be accepted.
2055 Default is to NOT accept these as it may be an un-intended
2060 - enabled if accept_ra_from_local is enabled
2061 on a specific interface.
2062 - disabled if accept_ra_from_local is disabled
2063 on a specific interface.
2065 accept_ra_min_hop_limit - INTEGER
2066 Minimum hop limit Information in Router Advertisement.
2068 Hop limit Information in Router Advertisement less than this
2069 variable shall be ignored.
2073 accept_ra_min_lft - INTEGER
2074 Minimum acceptable lifetime value in Router Advertisement.
2076 RA sections with a lifetime less than this value shall be
2077 ignored. Zero lifetimes stay unaffected.
2081 accept_ra_pinfo - BOOLEAN
2082 Learn Prefix Information in Router Advertisement.
2086 - enabled if accept_ra is enabled.
2087 - disabled if accept_ra is disabled.
2089 accept_ra_rt_info_min_plen - INTEGER
2090 Minimum prefix length of Route Information in RA.
2092 Route Information w/ prefix smaller than this variable shall
2097 * 0 if accept_ra_rtr_pref is enabled.
2098 * -1 if accept_ra_rtr_pref is disabled.
2100 accept_ra_rt_info_max_plen - INTEGER
2101 Maximum prefix length of Route Information in RA.
2103 Route Information w/ prefix larger than this variable shall
2108 * 0 if accept_ra_rtr_pref is enabled.
2109 * -1 if accept_ra_rtr_pref is disabled.
2111 accept_ra_rtr_pref - BOOLEAN
2112 Accept Router Preference in RA.
2116 - enabled if accept_ra is enabled.
2117 - disabled if accept_ra is disabled.
2119 accept_ra_mtu - BOOLEAN
2120 Apply the MTU value specified in RA option 5 (RFC4861). If
2121 disabled, the MTU specified in the RA will be ignored.
2125 - enabled if accept_ra is enabled.
2126 - disabled if accept_ra is disabled.
2128 accept_redirects - BOOLEAN
2133 - enabled if local forwarding is disabled.
2134 - disabled if local forwarding is enabled.
2136 accept_source_route - INTEGER
2137 Accept source routing (routing extension header).
2139 - >= 0: Accept only routing header type 2.
2140 - < 0: Do not accept routing header.
2145 Autoconfigure addresses using Prefix Information in Router
2150 - enabled if accept_ra_pinfo is enabled.
2151 - disabled if accept_ra_pinfo is disabled.
2153 dad_transmits - INTEGER
2154 The amount of Duplicate Address Detection probes to send.
2158 forwarding - INTEGER
2159 Configure interface-specific Host/Router behaviour.
2163 It is recommended to have the same setting on all
2164 interfaces; mixed router/host scenarios are rather uncommon.
2166 Possible values are:
2168 - 0 Forwarding disabled
2169 - 1 Forwarding enabled
2173 By default, Host behaviour is assumed. This means:
2175 1. IsRouter flag is not set in Neighbour Advertisements.
2176 2. If accept_ra is TRUE (default), transmit Router
2178 3. If accept_ra is TRUE (default), accept Router
2179 Advertisements (and do autoconfiguration).
2180 4. If accept_redirects is TRUE (default), accept Redirects.
2184 If local forwarding is enabled, Router behaviour is assumed.
2185 This means exactly the reverse from the above:
2187 1. IsRouter flag is set in Neighbour Advertisements.
2188 2. Router Solicitations are not sent unless accept_ra is 2.
2189 3. Router Advertisements are ignored unless accept_ra is 2.
2190 4. Redirects are ignored.
2192 Default: 0 (disabled) if global forwarding is disabled (default),
2193 otherwise 1 (enabled).
2196 Default Hop Limit to set.
2201 Default Maximum Transfer Unit
2203 Default: 1280 (IPv6 required minimum)
2205 ip_nonlocal_bind - BOOLEAN
2206 If set, allows processes to bind() to non-local IPv6 addresses,
2207 which can be quite useful - but may break some applications.
2211 router_probe_interval - INTEGER
2212 Minimum interval (in seconds) between Router Probing described
2217 router_solicitation_delay - INTEGER
2218 Number of seconds to wait after interface is brought up
2219 before sending Router Solicitations.
2223 router_solicitation_interval - INTEGER
2224 Number of seconds to wait between Router Solicitations.
2228 router_solicitations - INTEGER
2229 Number of Router Solicitations to send until assuming no
2230 routers are present.
2234 use_oif_addrs_only - BOOLEAN
2235 When enabled, the candidate source addresses for destinations
2236 routed via this interface are restricted to the set of addresses
2237 configured on this interface (vis. RFC 6724, section 4).
2241 use_tempaddr - INTEGER
2242 Preference for Privacy Extensions (RFC3041).
2244 * <= 0 : disable Privacy Extensions
2245 * == 1 : enable Privacy Extensions, but prefer public
2246 addresses over temporary addresses.
2247 * > 1 : enable Privacy Extensions and prefer temporary
2248 addresses over public addresses.
2252 * 0 (for most devices)
2253 * -1 (for point-to-point devices and loopback devices)
2255 temp_valid_lft - INTEGER
2256 valid lifetime (in seconds) for temporary addresses.
2258 Default: 172800 (2 days)
2260 temp_prefered_lft - INTEGER
2261 Preferred lifetime (in seconds) for temporary addresses.
2263 Default: 86400 (1 day)
2265 keep_addr_on_down - INTEGER
2266 Keep all IPv6 addresses on an interface down event. If set static
2267 global addresses with no expiration time are not flushed.
2270 * 0 : system default
2273 Default: 0 (addresses are removed)
2275 max_desync_factor - INTEGER
2276 Maximum value for DESYNC_FACTOR, which is a random value
2277 that ensures that clients don't synchronize with each
2278 other and generate new addresses at exactly the same time.
2279 value is in seconds.
2283 regen_max_retry - INTEGER
2284 Number of attempts before give up attempting to generate
2285 valid temporary addresses.
2289 max_addresses - INTEGER
2290 Maximum number of autoconfigured addresses per interface. Setting
2291 to zero disables the limitation. It is not recommended to set this
2292 value too large (or to zero) because it would be an easy way to
2293 crash the kernel by allowing too many addresses to be created.
2297 disable_ipv6 - BOOLEAN
2298 Disable IPv6 operation. If accept_dad is set to 2, this value
2299 will be dynamically set to TRUE if DAD fails for the link-local
2302 Default: FALSE (enable IPv6 operation)
2304 When this value is changed from 1 to 0 (IPv6 is being enabled),
2305 it will dynamically create a link-local address on the given
2306 interface and start Duplicate Address Detection, if necessary.
2308 When this value is changed from 0 to 1 (IPv6 is being disabled),
2309 it will dynamically delete all addresses and routes on the given
2310 interface. From now on it will not possible to add addresses/routes
2311 to the selected interface.
2313 accept_dad - INTEGER
2314 Whether to accept DAD (Duplicate Address Detection).
2316 == ==============================================================
2318 1 Enable DAD (default)
2319 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate
2320 link-local address has been found.
2321 == ==============================================================
2323 DAD operation and mode on a given interface will be selected according
2324 to the maximum value of conf/{all,interface}/accept_dad.
2326 force_tllao - BOOLEAN
2327 Enable sending the target link-layer address option even when
2328 responding to a unicast neighbor solicitation.
2332 Quoting from RFC 2461, section 4.4, Target link-layer address:
2334 "The option MUST be included for multicast solicitations in order to
2335 avoid infinite Neighbor Solicitation "recursion" when the peer node
2336 does not have a cache entry to return a Neighbor Advertisements
2337 message. When responding to unicast solicitations, the option can be
2338 omitted since the sender of the solicitation has the correct link-
2339 layer address; otherwise it would not have be able to send the unicast
2340 solicitation in the first place. However, including the link-layer
2341 address in this case adds little overhead and eliminates a potential
2342 race condition where the sender deletes the cached link-layer address
2343 prior to receiving a response to a previous solicitation."
2345 ndisc_notify - BOOLEAN
2346 Define mode for notification of address and device changes.
2348 * 0 - (default): do nothing
2349 * 1 - Generate unsolicited neighbour advertisements when device is brought
2350 up or hardware address changes.
2352 ndisc_tclass - INTEGER
2353 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
2354 Discovery (Router Solicitation, Router Advertisement, Neighbor
2355 Solicitation, Neighbor Advertisement, Redirect) messages.
2356 These 8 bits can be interpreted as 6 high order bits holding the DSCP
2357 value and 2 low order bits representing ECN (which you probably want
2362 mldv1_unsolicited_report_interval - INTEGER
2363 The interval in milliseconds in which the next unsolicited
2364 MLDv1 report retransmit will take place.
2366 Default: 10000 (10 seconds)
2368 mldv2_unsolicited_report_interval - INTEGER
2369 The interval in milliseconds in which the next unsolicited
2370 MLDv2 report retransmit will take place.
2372 Default: 1000 (1 second)
2374 force_mld_version - INTEGER
2375 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
2376 * 1 - Enforce to use MLD version 1
2377 * 2 - Enforce to use MLD version 2
2379 suppress_frag_ndisc - INTEGER
2380 Control RFC 6980 (Security Implications of IPv6 Fragmentation
2381 with IPv6 Neighbor Discovery) behavior:
2383 * 1 - (default) discard fragmented neighbor discovery packets
2384 * 0 - allow fragmented neighbor discovery packets
2386 optimistic_dad - BOOLEAN
2387 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
2389 * 0: disabled (default)
2392 Optimistic Duplicate Address Detection for the interface will be enabled
2393 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
2394 it will be disabled otherwise.
2396 use_optimistic - BOOLEAN
2397 If enabled, do not classify optimistic addresses as deprecated during
2398 source address selection. Preferred addresses will still be chosen
2399 before optimistic addresses, subject to other ranking in the source
2400 address selection algorithm.
2402 * 0: disabled (default)
2405 This will be enabled if at least one of
2406 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
2408 stable_secret - IPv6 address
2409 This IPv6 address will be used as a secret to generate IPv6
2410 addresses for link-local addresses and autoconfigured
2411 ones. All addresses generated after setting this secret will
2412 be stable privacy ones by default. This can be changed via the
2413 addrgenmode ip-link. conf/default/stable_secret is used as the
2414 secret for the namespace, the interface specific ones can
2415 overwrite that. Writes to conf/all/stable_secret are refused.
2417 It is recommended to generate this secret during installation
2418 of a system and keep it stable after that.
2420 By default the stable secret is unset.
2422 addr_gen_mode - INTEGER
2423 Defines how link-local and autoconf addresses are generated.
2425 = =================================================================
2426 0 generate address based on EUI64 (default)
2427 1 do no generate a link-local address, use EUI64 for addresses
2428 generated from autoconf
2429 2 generate stable privacy addresses, using the secret from
2430 stable_secret (RFC7217)
2431 3 generate stable privacy addresses, using a random secret if unset
2432 = =================================================================
2434 drop_unicast_in_l2_multicast - BOOLEAN
2435 Drop any unicast IPv6 packets that are received in link-layer
2436 multicast (or broadcast) frames.
2438 By default this is turned off.
2440 drop_unsolicited_na - BOOLEAN
2441 Drop all unsolicited neighbor advertisements, for example if there's
2442 a known good NA proxy on the network and such frames need not be used
2443 (or in the case of 802.11, must not be used to prevent attacks.)
2445 By default this is turned off.
2447 enhanced_dad - BOOLEAN
2448 Include a nonce option in the IPv6 neighbor solicitation messages used for
2449 duplicate address detection per RFC7527. A received DAD NS will only signal
2450 a duplicate address if the nonce is different. This avoids any false
2451 detection of duplicates due to loopback of the NS messages that we send.
2452 The nonce option will be sent on an interface unless both of
2453 conf/{all,interface}/enhanced_dad are set to FALSE.
2461 Limit the maximal rates for sending ICMPv6 messages.
2463 0 to disable any limiting,
2464 otherwise the minimal space between responses in milliseconds.
2468 ratemask - list of comma separated ranges
2469 For ICMPv6 message types matching the ranges in the ratemask, limit
2470 the sending of the message according to ratelimit parameter.
2472 The format used for both input and output is a comma separated
2473 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
2474 129). Writing to the file will clear all previous ranges of ICMPv6
2475 message types and update the current list with the input.
2477 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
2478 for numerical values of ICMPv6 message types, e.g. echo request is 128
2479 and echo reply is 129.
2481 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
2483 echo_ignore_all - BOOLEAN
2484 If set non-zero, then the kernel will ignore all ICMP ECHO
2485 requests sent to it over the IPv6 protocol.
2489 echo_ignore_multicast - BOOLEAN
2490 If set non-zero, then the kernel will ignore all ICMP ECHO
2491 requests sent to it over the IPv6 protocol via multicast.
2495 echo_ignore_anycast - BOOLEAN
2496 If set non-zero, then the kernel will ignore all ICMP ECHO
2497 requests sent to it over the IPv6 protocol destined to anycast address.
2501 xfrm6_gc_thresh - INTEGER
2502 (Obsolete since linux-4.14)
2503 The threshold at which we will start garbage collecting for IPv6
2504 destination cache entries. At twice this value the system will
2505 refuse new allocations.
2509 Pekka Savola <pekkas@netcore.fi>
2510 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2513 /proc/sys/net/bridge/* Variables:
2514 =================================
2516 bridge-nf-call-arptables - BOOLEAN
2517 - 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2522 bridge-nf-call-iptables - BOOLEAN
2523 - 1 : pass bridged IPv4 traffic to iptables' chains.
2528 bridge-nf-call-ip6tables - BOOLEAN
2529 - 1 : pass bridged IPv6 traffic to ip6tables' chains.
2534 bridge-nf-filter-vlan-tagged - BOOLEAN
2535 - 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2540 bridge-nf-filter-pppoe-tagged - BOOLEAN
2541 - 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2546 bridge-nf-pass-vlan-input-dev - BOOLEAN
2547 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2548 interface on the bridge and set the netfilter input device to the
2549 vlan. This allows use of e.g. "iptables -i br0.1" and makes the
2550 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no
2551 matching vlan interface is found, or this switch is off, the input
2552 device is set to the bridge interface.
2554 - 0: disable bridge netfilter vlan interface lookup.
2558 ``proc/sys/net/sctp/*`` Variables:
2559 ==================================
2561 addip_enable - BOOLEAN
2562 Enable or disable extension of Dynamic Address Reconfiguration
2563 (ADD-IP) functionality specified in RFC5061. This extension provides
2564 the ability to dynamically add and remove new addresses for the SCTP
2567 1: Enable extension.
2569 0: Disable extension.
2574 Enable or disable pf (pf is short for potentially failed) state. A value
2575 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2576 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2577 Since pf_retrans and path_max_retrans can be changed by userspace
2578 application, sometimes user expects to disable pf state by the value of
2579 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2580 or path_max_retrans is changed by the user application, this pf state is
2581 enabled. As such, it is necessary to add this to dynamically enable
2582 and disable pf state. See:
2583 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2593 Unset or enable/disable pf (pf is short for potentially failed) state
2594 exposure. Applications can control the exposure of the PF path state
2595 in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2596 sockopt. When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2597 SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2598 can be got via SCTP_GET_PEER_ADDR_INFO sockopt; When it's enabled,
2599 a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2600 SCTP_PF state and a SCTP_PF-state transport info can be got via
2601 SCTP_GET_PEER_ADDR_INFO sockopt; When it's diabled, no
2602 SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2603 trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2606 0: Unset pf state exposure, Compatible with old applications.
2608 1: Disable pf state exposure.
2610 2: Enable pf state exposure.
2614 addip_noauth_enable - BOOLEAN
2615 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2616 authentication to protect the operations of adding or removing new
2617 addresses. This requirement is mandated so that unauthorized hosts
2618 would not be able to hijack associations. However, older
2619 implementations may not have implemented this requirement while
2620 allowing the ADD-IP extension. For reasons of interoperability,
2621 we provide this variable to control the enforcement of the
2622 authentication requirement.
2624 == ===============================================================
2625 1 Allow ADD-IP extension to be used without authentication. This
2626 should only be set in a closed environment for interoperability
2627 with older implementations.
2629 0 Enforce the authentication requirement
2630 == ===============================================================
2634 auth_enable - BOOLEAN
2635 Enable or disable Authenticated Chunks extension. This extension
2636 provides the ability to send and receive authenticated chunks and is
2637 required for secure operation of Dynamic Address Reconfiguration
2640 - 1: Enable this extension.
2641 - 0: Disable this extension.
2645 prsctp_enable - BOOLEAN
2646 Enable or disable the Partial Reliability extension (RFC3758) which
2647 is used to notify peers that a given DATA should no longer be expected.
2649 - 1: Enable extension
2655 The limit of the number of new packets that can be initially sent. It
2656 controls how bursty the generated traffic can be.
2660 association_max_retrans - INTEGER
2661 Set the maximum number for retransmissions that an association can
2662 attempt deciding that the remote end is unreachable. If this value
2663 is exceeded, the association is terminated.
2667 max_init_retransmits - INTEGER
2668 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2669 that an association will attempt before declaring the destination
2670 unreachable and terminating.
2674 path_max_retrans - INTEGER
2675 The maximum number of retransmissions that will be attempted on a given
2676 path. Once this threshold is exceeded, the path is considered
2677 unreachable, and new traffic will use a different path when the
2678 association is multihomed.
2682 pf_retrans - INTEGER
2683 The number of retransmissions that will be attempted on a given path
2684 before traffic is redirected to an alternate transport (should one
2685 exist). Note this is distinct from path_max_retrans, as a path that
2686 passes the pf_retrans threshold can still be used. Its only
2687 deprioritized when a transmission path is selected by the stack. This
2688 setting is primarily used to enable fast failover mechanisms without
2689 having to reduce path_max_retrans to a very low value. See:
2690 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2691 for details. Note also that a value of pf_retrans > path_max_retrans
2692 disables this feature. Since both pf_retrans and path_max_retrans can
2693 be changed by userspace application, a variable pf_enable is used to
2698 ps_retrans - INTEGER
2699 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
2700 from section-5 "Primary Path Switchover" in rfc7829. The primary path
2701 will be changed to another active path when the path error counter on
2702 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
2703 to continue data transmission on a new working path even when the old
2704 primary destination address becomes active again". Note this feature
2705 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
2706 and its value can't be less than 'pf_retrans' when changing by sysctl.
2710 rto_initial - INTEGER
2711 The initial round trip timeout value in milliseconds that will be used
2712 in calculating round trip times. This is the initial time interval
2713 for retransmissions.
2718 The maximum value (in milliseconds) of the round trip timeout. This
2719 is the largest time interval that can elapse between retransmissions.
2724 The minimum value (in milliseconds) of the round trip timeout. This
2725 is the smallest time interval the can elapse between retransmissions.
2729 hb_interval - INTEGER
2730 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2731 are sent at the specified interval on idle paths to probe the state of
2732 a given path between 2 associations.
2736 sack_timeout - INTEGER
2737 The amount of time (in milliseconds) that the implementation will wait
2742 valid_cookie_life - INTEGER
2743 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2744 is used during association establishment.
2748 cookie_preserve_enable - BOOLEAN
2749 Enable or disable the ability to extend the lifetime of the SCTP cookie
2750 that is used during the establishment phase of SCTP association
2752 - 1: Enable cookie lifetime extension.
2757 cookie_hmac_alg - STRING
2758 Select the hmac algorithm used when generating the cookie value sent by
2759 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2766 Ability to assign md5 or sha1 as the selected alg is predicated on the
2767 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2768 CONFIG_CRYPTO_SHA1).
2770 Default: Dependent on configuration. MD5 if available, else SHA1 if
2771 available, else none.
2773 rcvbuf_policy - INTEGER
2774 Determines if the receive buffer is attributed to the socket or to
2775 association. SCTP supports the capability to create multiple
2776 associations on a single socket. When using this capability, it is
2777 possible that a single stalled association that's buffering a lot
2778 of data may block other associations from delivering their data by
2779 consuming all of the receive buffer space. To work around this,
2780 the rcvbuf_policy could be set to attribute the receiver buffer space
2781 to each association instead of the socket. This prevents the described
2784 - 1: rcvbuf space is per association
2785 - 0: rcvbuf space is per socket
2789 sndbuf_policy - INTEGER
2790 Similar to rcvbuf_policy above, this applies to send buffer space.
2792 - 1: Send buffer is tracked per association
2793 - 0: Send buffer is tracked per socket.
2797 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2798 Number of pages allowed for queueing by all SCTP sockets.
2800 min: Below this number of pages SCTP is not bothered about its
2801 memory appetite. When amount of memory allocated by SCTP exceeds
2802 this number, SCTP starts to moderate memory usage.
2804 pressure: This value was introduced to follow format of tcp_mem.
2806 max: Number of pages allowed for queueing by all SCTP sockets.
2808 Default is calculated at boot time from amount of available memory.
2810 sctp_rmem - vector of 3 INTEGERs: min, default, max
2811 Only the first value ("min") is used, "default" and "max" are
2814 min: Minimal size of receive buffer used by SCTP socket.
2815 It is guaranteed to each SCTP socket (but not association) even
2816 under moderate memory pressure.
2820 sctp_wmem - vector of 3 INTEGERs: min, default, max
2821 Only the first value ("min") is used, "default" and "max" are
2824 min: Minimum size of send buffer that can be used by SCTP sockets.
2825 It is guaranteed to each SCTP socket (but not association) even
2826 under moderate memory pressure.
2830 addr_scope_policy - INTEGER
2831 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2833 - 0 - Disable IPv4 address scoping
2834 - 1 - Enable IPv4 address scoping
2835 - 2 - Follow draft but allow IPv4 private addresses
2836 - 3 - Follow draft but allow IPv4 link local addresses
2841 The listening port for the local UDP tunneling sock. Normally it's
2842 using the IANA-assigned UDP port number 9899 (sctp-tunneling).
2844 This UDP sock is used for processing the incoming UDP-encapsulated
2845 SCTP packets (from RFC6951), and shared by all applications in the
2846 same net namespace. This UDP sock will be closed when the value is
2849 The value will also be used to set the src port of the UDP header
2850 for the outgoing UDP-encapsulated SCTP packets. For the dest port,
2851 please refer to 'encap_port' below.
2855 encap_port - INTEGER
2856 The default remote UDP encapsulation port.
2858 This value is used to set the dest port of the UDP header for the
2859 outgoing UDP-encapsulated SCTP packets by default. Users can also
2860 change the value for each sock/asoc/transport by using setsockopt.
2861 For further information, please refer to RFC6951.
2863 Note that when connecting to a remote server, the client should set
2864 this to the port that the UDP tunneling sock on the peer server is
2865 listening to and the local UDP tunneling sock on the client also
2866 must be started. On the server, it would get the encap_port from
2867 the incoming packet's source port.
2871 plpmtud_probe_interval - INTEGER
2872 The time interval (in milliseconds) for the PLPMTUD probe timer,
2873 which is configured to expire after this period to receive an
2874 acknowledgment to a probe packet. This is also the time interval
2875 between the probes for the current pmtu when the probe search
2878 PLPMTUD will be disabled when 0 is set, and other values for it
2884 ``/proc/sys/net/core/*``
2885 ========================
2887 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
2890 ``/proc/sys/net/unix/*``
2891 ========================
2893 max_dgram_qlen - INTEGER
2894 The maximum length of dgram socket receive queue