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 the smallest of the old MTU to
29 this destination and min_pmtu (see below). You will need
30 to raise min_pmtu to the smallest interface MTU on your system
31 manually if you want to avoid locally generated fragments.
33 In mode 2 incoming Path MTU Discovery messages will be
34 discarded. Outgoing frames are handled the same as in mode 1,
35 implicitly setting IP_PMTUDISC_DONT on every created socket.
37 Mode 3 is a hardened pmtu discover mode. The kernel will only
38 accept fragmentation-needed errors if the underlying protocol
39 can verify them besides a plain socket lookup. Current
40 protocols for which pmtu events will be honored are TCP, SCTP
41 and DCCP as they verify e.g. the sequence number or the
42 association. This mode should not be enabled globally but is
43 only intended to secure e.g. name servers in namespaces where
44 TCP path mtu must still work but path MTU information of other
45 protocols should be discarded. If enabled globally this mode
46 could break other protocols.
53 default 552 - minimum Path MTU. Unless this is changed mannually,
54 each cached pmtu will never be lower than this setting.
56 ip_forward_use_pmtu - BOOLEAN
57 By default we don't trust protocol path MTUs while forwarding
58 because they could be easily forged and can lead to unwanted
59 fragmentation by the router.
60 You only need to enable this if you have user-space software
61 which tries to discover path mtus by itself and depends on the
62 kernel honoring this information. This is normally not the
72 fwmark_reflect - BOOLEAN
73 Controls the fwmark of kernel-generated IPv4 reply packets that are not
74 associated with a socket for example, TCP RSTs or ICMP echo replies).
75 If unset, these packets have a fwmark of zero. If set, they have the
76 fwmark of the packet they are replying to.
80 fib_multipath_use_neigh - BOOLEAN
81 Use status of existing neighbor entry when determining nexthop for
82 multipath routes. If disabled, neighbor information is not used and
83 packets could be directed to a failed nexthop. Only valid for kernels
84 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
93 fib_multipath_hash_policy - INTEGER
94 Controls which hash policy to use for multipath routes. Only valid
95 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
103 - 2 - Layer 3 or inner Layer 3 if present
104 - 3 - Custom multipath hash. Fields used for multipath hash calculation
105 are determined by fib_multipath_hash_fields sysctl
107 fib_multipath_hash_fields - UNSIGNED INTEGER
108 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
109 fields used for multipath hash calculation are determined by this
112 This value is a bitmask which enables various fields for multipath hash
117 ====== ============================
118 0x0001 Source IP address
119 0x0002 Destination IP address
121 0x0008 Unused (Flow Label)
123 0x0020 Destination port
124 0x0040 Inner source IP address
125 0x0080 Inner destination IP address
126 0x0100 Inner IP protocol
127 0x0200 Inner Flow Label
128 0x0400 Inner source port
129 0x0800 Inner destination port
130 ====== ============================
132 Default: 0x0007 (source IP, destination IP and IP protocol)
134 fib_sync_mem - UNSIGNED INTEGER
135 Amount of dirty memory from fib entries that can be backlogged before
136 synchronize_rcu is forced.
138 Default: 512kB Minimum: 64kB Maximum: 64MB
140 ip_forward_update_priority - INTEGER
141 Whether to update SKB priority from "TOS" field in IPv4 header after it
142 is forwarded. The new SKB priority is mapped from TOS field value
143 according to an rt_tos2priority table (see e.g. man tc-prio).
145 Default: 1 (Update priority.)
149 - 0 - Do not update priority.
150 - 1 - Update priority.
152 route/max_size - INTEGER
153 Maximum number of routes allowed in the kernel. Increase
154 this when using large numbers of interfaces and/or routes.
156 From linux kernel 3.6 onwards, this is deprecated for ipv4
157 as route cache is no longer used.
159 neigh/default/gc_thresh1 - INTEGER
160 Minimum number of entries to keep. Garbage collector will not
161 purge entries if there are fewer than this number.
165 neigh/default/gc_thresh2 - INTEGER
166 Threshold when garbage collector becomes more aggressive about
167 purging entries. Entries older than 5 seconds will be cleared
168 when over this number.
172 neigh/default/gc_thresh3 - INTEGER
173 Maximum number of non-PERMANENT neighbor entries allowed. Increase
174 this when using large numbers of interfaces and when communicating
175 with large numbers of directly-connected peers.
179 neigh/default/unres_qlen_bytes - INTEGER
180 The maximum number of bytes which may be used by packets
181 queued for each unresolved address by other network layers.
184 Setting negative value is meaningless and will return error.
186 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
188 Exact value depends on architecture and kernel options,
189 but should be enough to allow queuing 256 packets
192 neigh/default/unres_qlen - INTEGER
193 The maximum number of packets which may be queued for each
194 unresolved address by other network layers.
196 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
198 Prior to linux 3.3, the default value is 3 which may cause
199 unexpected packet loss. The current default value is calculated
200 according to default value of unres_qlen_bytes and true size of
205 mtu_expires - INTEGER
206 Time, in seconds, that cached PMTU information is kept.
208 min_adv_mss - INTEGER
209 The advertised MSS depends on the first hop route MTU, but will
210 never be lower than this setting.
212 fib_notify_on_flag_change - INTEGER
213 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
214 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
216 After installing a route to the kernel, user space receives an
217 acknowledgment, which means the route was installed in the kernel,
218 but not necessarily in hardware.
219 It is also possible for a route already installed in hardware to change
220 its action and therefore its flags. For example, a host route that is
221 trapping packets can be "promoted" to perform decapsulation following
222 the installation of an IPinIP/VXLAN tunnel.
223 The notifications will indicate to user-space the state of the route.
225 Default: 0 (Do not emit notifications.)
229 - 0 - Do not emit notifications.
230 - 1 - Emit notifications.
231 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
235 ipfrag_high_thresh - LONG INTEGER
236 Maximum memory used to reassemble IP fragments.
238 ipfrag_low_thresh - LONG INTEGER
239 (Obsolete since linux-4.17)
240 Maximum memory used to reassemble IP fragments before the kernel
241 begins to remove incomplete fragment queues to free up resources.
242 The kernel still accepts new fragments for defragmentation.
244 ipfrag_time - INTEGER
245 Time in seconds to keep an IP fragment in memory.
247 ipfrag_max_dist - INTEGER
248 ipfrag_max_dist is a non-negative integer value which defines the
249 maximum "disorder" which is allowed among fragments which share a
250 common IP source address. Note that reordering of packets is
251 not unusual, but if a large number of fragments arrive from a source
252 IP address while a particular fragment queue remains incomplete, it
253 probably indicates that one or more fragments belonging to that queue
254 have been lost. When ipfrag_max_dist is positive, an additional check
255 is done on fragments before they are added to a reassembly queue - if
256 ipfrag_max_dist (or more) fragments have arrived from a particular IP
257 address between additions to any IP fragment queue using that source
258 address, it's presumed that one or more fragments in the queue are
259 lost. The existing fragment queue will be dropped, and a new one
260 started. An ipfrag_max_dist value of zero disables this check.
262 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
263 result in unnecessarily dropping fragment queues when normal
264 reordering of packets occurs, which could lead to poor application
265 performance. Using a very large value, e.g. 50000, increases the
266 likelihood of incorrectly reassembling IP fragments that originate
267 from different IP datagrams, which could result in data corruption.
270 bc_forwarding - INTEGER
271 bc_forwarding enables the feature described in rfc1812#section-5.3.5.2
272 and rfc2644. It allows the router to forward directed broadcast.
273 To enable this feature, the 'all' entry and the input interface entry
280 inet_peer_threshold - INTEGER
281 The approximate size of the storage. Starting from this threshold
282 entries will be thrown aggressively. This threshold also determines
283 entries' time-to-live and time intervals between garbage collection
284 passes. More entries, less time-to-live, less GC interval.
286 inet_peer_minttl - INTEGER
287 Minimum time-to-live of entries. Should be enough to cover fragment
288 time-to-live on the reassembling side. This minimum time-to-live is
289 guaranteed if the pool size is less than inet_peer_threshold.
292 inet_peer_maxttl - INTEGER
293 Maximum time-to-live of entries. Unused entries will expire after
294 this period of time if there is no memory pressure on the pool (i.e.
295 when the number of entries in the pool is very small).
302 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
303 Defaults to 4096. (Was 128 before linux-5.4)
304 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
306 tcp_abort_on_overflow - BOOLEAN
307 If listening service is too slow to accept new connections,
308 reset them. Default state is FALSE. It means that if overflow
309 occurred due to a burst, connection will recover. Enable this
310 option _only_ if you are really sure that listening daemon
311 cannot be tuned to accept connections faster. Enabling this
312 option can harm clients of your server.
314 tcp_adv_win_scale - INTEGER
315 Count buffering overhead as bytes/2^tcp_adv_win_scale
316 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
319 Possible values are [-31, 31], inclusive.
323 tcp_allowed_congestion_control - STRING
324 Show/set the congestion control choices available to non-privileged
325 processes. The list is a subset of those listed in
326 tcp_available_congestion_control.
328 Default is "reno" and the default setting (tcp_congestion_control).
330 tcp_app_win - INTEGER
331 Reserve max(window/2^tcp_app_win, mss) of window for application
332 buffer. Value 0 is special, it means that nothing is reserved.
336 tcp_autocorking - BOOLEAN
337 Enable TCP auto corking :
338 When applications do consecutive small write()/sendmsg() system calls,
339 we try to coalesce these small writes as much as possible, to lower
340 total amount of sent packets. This is done if at least one prior
341 packet for the flow is waiting in Qdisc queues or device transmit
342 queue. Applications can still use TCP_CORK for optimal behavior
343 when they know how/when to uncork their sockets.
347 tcp_available_congestion_control - STRING
348 Shows the available congestion control choices that are registered.
349 More congestion control algorithms may be available as modules,
352 tcp_base_mss - INTEGER
353 The initial value of search_low to be used by the packetization layer
354 Path MTU discovery (MTU probing). If MTU probing is enabled,
355 this is the initial MSS used by the connection.
357 tcp_mtu_probe_floor - INTEGER
358 If MTU probing is enabled this caps the minimum MSS used for search_low
363 tcp_min_snd_mss - INTEGER
364 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
365 as described in RFC 1122 and RFC 6691.
367 If this ADVMSS option is smaller than tcp_min_snd_mss,
368 it is silently capped to tcp_min_snd_mss.
370 Default : 48 (at least 8 bytes of payload per segment)
372 tcp_congestion_control - STRING
373 Set the congestion control algorithm to be used for new
374 connections. The algorithm "reno" is always available, but
375 additional choices may be available based on kernel configuration.
376 Default is set as part of kernel configuration.
377 For passive connections, the listener congestion control choice
380 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
383 Allows TCP to send "duplicate" SACKs.
385 tcp_early_retrans - INTEGER
386 Tail loss probe (TLP) converts RTOs occurring due to tail
387 losses into fast recovery (draft-ietf-tcpm-rack). Note that
388 TLP requires RACK to function properly (see tcp_recovery below)
398 Control use of Explicit Congestion Notification (ECN) by TCP.
399 ECN is used only when both ends of the TCP connection indicate
400 support for it. This feature is useful in avoiding losses due
401 to congestion by allowing supporting routers to signal
402 congestion before having to drop packets.
406 = =====================================================
407 0 Disable ECN. Neither initiate nor accept ECN.
408 1 Enable ECN when requested by incoming connections and
409 also request ECN on outgoing connection attempts.
410 2 Enable ECN when requested by incoming connections
411 but do not request ECN on outgoing connections.
412 = =====================================================
416 tcp_ecn_fallback - BOOLEAN
417 If the kernel detects that ECN connection misbehaves, enable fall
418 back to non-ECN. Currently, this knob implements the fallback
419 from RFC3168, section 6.1.1.1., but we reserve that in future,
420 additional detection mechanisms could be implemented under this
421 knob. The value is not used, if tcp_ecn or per route (or congestion
422 control) ECN settings are disabled.
424 Default: 1 (fallback enabled)
427 This is a legacy option, it has no effect anymore.
429 tcp_fin_timeout - INTEGER
430 The length of time an orphaned (no longer referenced by any
431 application) connection will remain in the FIN_WAIT_2 state
432 before it is aborted at the local end. While a perfectly
433 valid "receive only" state for an un-orphaned connection, an
434 orphaned connection in FIN_WAIT_2 state could otherwise wait
435 forever for the remote to close its end of the connection.
442 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
443 F-RTO is an enhanced recovery algorithm for TCP retransmission
444 timeouts. It is particularly beneficial in networks where the
445 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
446 modification. It does not require any support from the peer.
448 By default it's enabled with a non-zero value. 0 disables F-RTO.
450 tcp_fwmark_accept - BOOLEAN
451 If set, incoming connections to listening sockets that do not have a
452 socket mark will set the mark of the accepting socket to the fwmark of
453 the incoming SYN packet. This will cause all packets on that connection
454 (starting from the first SYNACK) to be sent with that fwmark. The
455 listening socket's mark is unchanged. Listening sockets that already
456 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
461 tcp_invalid_ratelimit - INTEGER
462 Limit the maximal rate for sending duplicate acknowledgments
463 in response to incoming TCP packets that are for an existing
464 connection but that are invalid due to any of these reasons:
466 (a) out-of-window sequence number,
467 (b) out-of-window acknowledgment number, or
468 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
470 This can help mitigate simple "ack loop" DoS attacks, wherein
471 a buggy or malicious middlebox or man-in-the-middle can
472 rewrite TCP header fields in manner that causes each endpoint
473 to think that the other is sending invalid TCP segments, thus
474 causing each side to send an unterminating stream of duplicate
475 acknowledgments for invalid segments.
477 Using 0 disables rate-limiting of dupacks in response to
478 invalid segments; otherwise this value specifies the minimal
479 space between sending such dupacks, in milliseconds.
481 Default: 500 (milliseconds).
483 tcp_keepalive_time - INTEGER
484 How often TCP sends out keepalive messages when keepalive is enabled.
487 tcp_keepalive_probes - INTEGER
488 How many keepalive probes TCP sends out, until it decides that the
489 connection is broken. Default value: 9.
491 tcp_keepalive_intvl - INTEGER
492 How frequently the probes are send out. Multiplied by
493 tcp_keepalive_probes it is time to kill not responding connection,
494 after probes started. Default value: 75sec i.e. connection
495 will be aborted after ~11 minutes of retries.
497 tcp_l3mdev_accept - BOOLEAN
498 Enables child sockets to inherit the L3 master device index.
499 Enabling this option allows a "global" listen socket to work
500 across L3 master domains (e.g., VRFs) with connected sockets
501 derived from the listen socket to be bound to the L3 domain in
502 which the packets originated. Only valid when the kernel was
503 compiled with CONFIG_NET_L3_MASTER_DEV.
505 Default: 0 (disabled)
507 tcp_low_latency - BOOLEAN
508 This is a legacy option, it has no effect anymore.
510 tcp_max_orphans - INTEGER
511 Maximal number of TCP sockets not attached to any user file handle,
512 held by system. If this number is exceeded orphaned connections are
513 reset immediately and warning is printed. This limit exists
514 only to prevent simple DoS attacks, you _must_ not rely on this
515 or lower the limit artificially, but rather increase it
516 (probably, after increasing installed memory),
517 if network conditions require more than default value,
518 and tune network services to linger and kill such states
519 more aggressively. Let me to remind again: each orphan eats
520 up to ~64K of unswappable memory.
522 tcp_max_syn_backlog - INTEGER
523 Maximal number of remembered connection requests (SYN_RECV),
524 which have not received an acknowledgment from connecting client.
526 This is a per-listener limit.
528 The minimal value is 128 for low memory machines, and it will
529 increase in proportion to the memory of machine.
531 If server suffers from overload, try increasing this number.
533 Remember to also check /proc/sys/net/core/somaxconn
534 A SYN_RECV request socket consumes about 304 bytes of memory.
536 tcp_max_tw_buckets - INTEGER
537 Maximal number of timewait sockets held by system simultaneously.
538 If this number is exceeded time-wait socket is immediately destroyed
539 and warning is printed. This limit exists only to prevent
540 simple DoS attacks, you _must_ not lower the limit artificially,
541 but rather increase it (probably, after increasing installed memory),
542 if network conditions require more than default value.
544 tcp_mem - vector of 3 INTEGERs: min, pressure, max
545 min: below this number of pages TCP is not bothered about its
548 pressure: when amount of memory allocated by TCP exceeds this number
549 of pages, TCP moderates its memory consumption and enters memory
550 pressure mode, which is exited when memory consumption falls
553 max: number of pages allowed for queueing by all TCP sockets.
555 Defaults are calculated at boot time from amount of available
558 tcp_min_rtt_wlen - INTEGER
559 The window length of the windowed min filter to track the minimum RTT.
560 A shorter window lets a flow more quickly pick up new (higher)
561 minimum RTT when it is moved to a longer path (e.g., due to traffic
562 engineering). A longer window makes the filter more resistant to RTT
563 inflations such as transient congestion. The unit is seconds.
565 Possible values: 0 - 86400 (1 day)
569 tcp_moderate_rcvbuf - BOOLEAN
570 If set, TCP performs receive buffer auto-tuning, attempting to
571 automatically size the buffer (no greater than tcp_rmem[2]) to
572 match the size required by the path for full throughput. Enabled by
575 tcp_mtu_probing - INTEGER
576 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
580 - 1 - Disabled by default, enabled when an ICMP black hole detected
581 - 2 - Always enabled, use initial MSS of tcp_base_mss.
583 tcp_probe_interval - UNSIGNED INTEGER
584 Controls how often to start TCP Packetization-Layer Path MTU
585 Discovery reprobe. The default is reprobing every 10 minutes as
588 tcp_probe_threshold - INTEGER
589 Controls when TCP Packetization-Layer Path MTU Discovery probing
590 will stop in respect to the width of search range in bytes. Default
593 tcp_no_metrics_save - BOOLEAN
594 By default, TCP saves various connection metrics in the route cache
595 when the connection closes, so that connections established in the
596 near future can use these to set initial conditions. Usually, this
597 increases overall performance, but may sometimes cause performance
598 degradation. If set, TCP will not cache metrics on closing
601 tcp_no_ssthresh_metrics_save - BOOLEAN
602 Controls whether TCP saves ssthresh metrics in the route cache.
604 Default is 1, which disables ssthresh metrics.
606 tcp_orphan_retries - INTEGER
607 This value influences the timeout of a locally closed TCP connection,
608 when RTO retransmissions remain unacknowledged.
609 See tcp_retries2 for more details.
611 The default value is 8.
613 If your machine is a loaded WEB server,
614 you should think about lowering this value, such sockets
615 may consume significant resources. Cf. tcp_max_orphans.
617 tcp_recovery - INTEGER
618 This value is a bitmap to enable various experimental loss recovery
621 ========= =============================================================
622 RACK: 0x1 enables the RACK loss detection for fast detection of lost
623 retransmissions and tail drops. It also subsumes and disables
624 RFC6675 recovery for SACK connections.
626 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
628 RACK: 0x4 disables RACK's DUPACK threshold heuristic
629 ========= =============================================================
633 tcp_reordering - INTEGER
634 Initial reordering level of packets in a TCP stream.
635 TCP stack can then dynamically adjust flow reordering level
636 between this initial value and tcp_max_reordering
640 tcp_max_reordering - INTEGER
641 Maximal reordering level of packets in a TCP stream.
642 300 is a fairly conservative value, but you might increase it
643 if paths are using per packet load balancing (like bonding rr mode)
647 tcp_retrans_collapse - BOOLEAN
648 Bug-to-bug compatibility with some broken printers.
649 On retransmit try to send bigger packets to work around bugs in
652 tcp_retries1 - INTEGER
653 This value influences the time, after which TCP decides, that
654 something is wrong due to unacknowledged RTO retransmissions,
655 and reports this suspicion to the network layer.
656 See tcp_retries2 for more details.
658 RFC 1122 recommends at least 3 retransmissions, which is the
661 tcp_retries2 - INTEGER
662 This value influences the timeout of an alive TCP connection,
663 when RTO retransmissions remain unacknowledged.
664 Given a value of N, a hypothetical TCP connection following
665 exponential backoff with an initial RTO of TCP_RTO_MIN would
666 retransmit N times before killing the connection at the (N+1)th RTO.
668 The default value of 15 yields a hypothetical timeout of 924.6
669 seconds and is a lower bound for the effective timeout.
670 TCP will effectively time out at the first RTO which exceeds the
671 hypothetical timeout.
673 RFC 1122 recommends at least 100 seconds for the timeout,
674 which corresponds to a value of at least 8.
676 tcp_rfc1337 - BOOLEAN
677 If set, the TCP stack behaves conforming to RFC1337. If unset,
678 we are not conforming to RFC, but prevent TCP TIME_WAIT
683 tcp_rmem - vector of 3 INTEGERs: min, default, max
684 min: Minimal size of receive buffer used by TCP sockets.
685 It is guaranteed to each TCP socket, even under moderate memory
690 default: initial size of receive buffer used by TCP sockets.
691 This value overrides net.core.rmem_default used by other protocols.
692 Default: 131072 bytes.
693 This value results in initial window of 65535.
695 max: maximal size of receive buffer allowed for automatically
696 selected receiver buffers for TCP socket. This value does not override
697 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
698 automatic tuning of that socket's receive buffer size, in which
699 case this value is ignored.
700 Default: between 131072 and 6MB, depending on RAM size.
703 Enable select acknowledgments (SACKS).
705 tcp_comp_sack_delay_ns - LONG INTEGER
706 TCP tries to reduce number of SACK sent, using a timer
707 based on 5% of SRTT, capped by this sysctl, in nano seconds.
708 The default is 1ms, based on TSO autosizing period.
710 Default : 1,000,000 ns (1 ms)
712 tcp_comp_sack_slack_ns - LONG INTEGER
713 This sysctl control the slack used when arming the
714 timer used by SACK compression. This gives extra time
715 for small RTT flows, and reduces system overhead by allowing
716 opportunistic reduction of timer interrupts.
718 Default : 100,000 ns (100 us)
720 tcp_comp_sack_nr - INTEGER
721 Max number of SACK that can be compressed.
722 Using 0 disables SACK compression.
726 tcp_slow_start_after_idle - BOOLEAN
727 If set, provide RFC2861 behavior and time out the congestion
728 window after an idle period. An idle period is defined at
729 the current RTO. If unset, the congestion window will not
730 be timed out after an idle period.
735 Use the Host requirements interpretation of the TCP urgent pointer field.
736 Most hosts use the older BSD interpretation, so if you turn this on
737 Linux might not communicate correctly with them.
741 tcp_synack_retries - INTEGER
742 Number of times SYNACKs for a passive TCP connection attempt will
743 be retransmitted. Should not be higher than 255. Default value
744 is 5, which corresponds to 31seconds till the last retransmission
745 with the current initial RTO of 1second. With this the final timeout
746 for a passive TCP connection will happen after 63seconds.
748 tcp_syncookies - INTEGER
749 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
750 Send out syncookies when the syn backlog queue of a socket
751 overflows. This is to prevent against the common 'SYN flood attack'
754 Note, that syncookies is fallback facility.
755 It MUST NOT be used to help highly loaded servers to stand
756 against legal connection rate. If you see SYN flood warnings
757 in your logs, but investigation shows that they occur
758 because of overload with legal connections, you should tune
759 another parameters until this warning disappear.
760 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
762 syncookies seriously violate TCP protocol, do not allow
763 to use TCP extensions, can result in serious degradation
764 of some services (f.e. SMTP relaying), visible not by you,
765 but your clients and relays, contacting you. While you see
766 SYN flood warnings in logs not being really flooded, your server
767 is seriously misconfigured.
769 If you want to test which effects syncookies have to your
770 network connections you can set this knob to 2 to enable
771 unconditionally generation of syncookies.
773 tcp_migrate_req - BOOLEAN
774 The incoming connection is tied to a specific listening socket when
775 the initial SYN packet is received during the three-way handshake.
776 When a listener is closed, in-flight request sockets during the
777 handshake and established sockets in the accept queue are aborted.
779 If the listener has SO_REUSEPORT enabled, other listeners on the
780 same port should have been able to accept such connections. This
781 option makes it possible to migrate such child sockets to another
782 listener after close() or shutdown().
784 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
785 usually be used to define the policy to pick an alive listener.
786 Otherwise, the kernel will randomly pick an alive listener only if
787 this option is enabled.
789 Note that migration between listeners with different settings may
790 crash applications. Let's say migration happens from listener A to
791 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
792 the requests migrated from A. To avoid such a situation, cancel
793 migration by returning SK_DROP in the type of eBPF program, or
798 tcp_fastopen - INTEGER
799 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
802 The client support is enabled by flag 0x1 (on by default). The client
803 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
804 rather than connect() to send data in SYN.
806 The server support is enabled by flag 0x2 (off by default). Then
807 either enable for all listeners with another flag (0x400) or
808 enable individual listeners via TCP_FASTOPEN socket option with
809 the option value being the length of the syn-data backlog.
811 The values (bitmap) are
813 ===== ======== ======================================================
814 0x1 (client) enables sending data in the opening SYN on the client.
815 0x2 (server) enables the server support, i.e., allowing data in
816 a SYN packet to be accepted and passed to the
817 application before 3-way handshake finishes.
818 0x4 (client) send data in the opening SYN regardless of cookie
819 availability and without a cookie option.
820 0x200 (server) accept data-in-SYN w/o any cookie option present.
821 0x400 (server) enable all listeners to support Fast Open by
822 default without explicit TCP_FASTOPEN socket option.
823 ===== ======== ======================================================
827 Note that additional client or server features are only
828 effective if the basic support (0x1 and 0x2) are enabled respectively.
830 tcp_fastopen_blackhole_timeout_sec - INTEGER
831 Initial time period in second to disable Fastopen on active TCP sockets
832 when a TFO firewall blackhole issue happens.
833 This time period will grow exponentially when more blackhole issues
834 get detected right after Fastopen is re-enabled and will reset to
835 initial value when the blackhole issue goes away.
836 0 to disable the blackhole detection.
838 By default, it is set to 0 (feature is disabled).
840 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
841 The list consists of a primary key and an optional backup key. The
842 primary key is used for both creating and validating cookies, while the
843 optional backup key is only used for validating cookies. The purpose of
844 the backup key is to maximize TFO validation when keys are rotated.
846 A randomly chosen primary key may be configured by the kernel if
847 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
848 TCP_FASTOPEN setsockopt() optname is set and a key has not been
849 previously configured via sysctl. If keys are configured via
850 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
851 per-socket keys will be used instead of any keys that are specified via
854 A key is specified as 4 8-digit hexadecimal integers which are separated
855 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
856 omitted. A primary and a backup key may be specified by separating them
857 by a comma. If only one key is specified, it becomes the primary key and
858 any previously configured backup keys are removed.
860 tcp_syn_retries - INTEGER
861 Number of times initial SYNs for an active TCP connection attempt
862 will be retransmitted. Should not be higher than 127. Default value
863 is 6, which corresponds to 63seconds till the last retransmission
864 with the current initial RTO of 1second. With this the final timeout
865 for an active TCP connection attempt will happen after 127seconds.
867 tcp_timestamps - INTEGER
868 Enable timestamps as defined in RFC1323.
871 - 1: Enable timestamps as defined in RFC1323 and use random offset for
872 each connection rather than only using the current time.
873 - 2: Like 1, but without random offsets.
877 tcp_min_tso_segs - INTEGER
878 Minimal number of segments per TSO frame.
880 Since linux-3.12, TCP does an automatic sizing of TSO frames,
881 depending on flow rate, instead of filling 64Kbytes packets.
882 For specific usages, it's possible to force TCP to build big
883 TSO frames. Note that TCP stack might split too big TSO packets
884 if available window is too small.
888 tcp_tso_rtt_log - INTEGER
889 Adjustment of TSO packet sizes based on min_rtt
891 Starting from linux-5.18, TCP autosizing can be tweaked
892 for flows having small RTT.
894 Old autosizing was splitting the pacing budget to send 1024 TSO
897 tso_packet_size = sk->sk_pacing_rate / 1024;
899 With the new mechanism, we increase this TSO sizing using:
901 distance = min_rtt_usec / (2^tcp_tso_rtt_log)
902 tso_packet_size += gso_max_size >> distance;
904 This means that flows between very close hosts can use bigger
905 TSO packets, reducing their cpu costs.
907 If you want to use the old autosizing, set this sysctl to 0.
909 Default: 9 (2^9 = 512 usec)
911 tcp_pacing_ss_ratio - INTEGER
912 sk->sk_pacing_rate is set by TCP stack using a ratio applied
913 to current rate. (current_rate = cwnd * mss / srtt)
914 If TCP is in slow start, tcp_pacing_ss_ratio is applied
915 to let TCP probe for bigger speeds, assuming cwnd can be
916 doubled every other RTT.
920 tcp_pacing_ca_ratio - INTEGER
921 sk->sk_pacing_rate is set by TCP stack using a ratio applied
922 to current rate. (current_rate = cwnd * mss / srtt)
923 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
924 is applied to conservatively probe for bigger throughput.
928 tcp_tso_win_divisor - INTEGER
929 This allows control over what percentage of the congestion window
930 can be consumed by a single TSO frame.
931 The setting of this parameter is a choice between burstiness and
932 building larger TSO frames.
936 tcp_tw_reuse - INTEGER
937 Enable reuse of TIME-WAIT sockets for new connections when it is
938 safe from protocol viewpoint.
942 - 2 - enable for loopback traffic only
944 It should not be changed without advice/request of technical
949 tcp_window_scaling - BOOLEAN
950 Enable window scaling as defined in RFC1323.
952 tcp_wmem - vector of 3 INTEGERs: min, default, max
953 min: Amount of memory reserved for send buffers for TCP sockets.
954 Each TCP socket has rights to use it due to fact of its birth.
958 default: initial size of send buffer used by TCP sockets. This
959 value overrides net.core.wmem_default used by other protocols.
961 It is usually lower than net.core.wmem_default.
965 max: Maximal amount of memory allowed for automatically tuned
966 send buffers for TCP sockets. This value does not override
967 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
968 automatic tuning of that socket's send buffer size, in which case
969 this value is ignored.
971 Default: between 64K and 4MB, depending on RAM size.
973 tcp_notsent_lowat - UNSIGNED INTEGER
974 A TCP socket can control the amount of unsent bytes in its write queue,
975 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
976 reports POLLOUT events if the amount of unsent bytes is below a per
977 socket value, and if the write queue is not full. sendmsg() will
978 also not add new buffers if the limit is hit.
980 This global variable controls the amount of unsent data for
981 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
982 to the global variable has immediate effect.
984 Default: UINT_MAX (0xFFFFFFFF)
986 tcp_workaround_signed_windows - BOOLEAN
987 If set, assume no receipt of a window scaling option means the
988 remote TCP is broken and treats the window as a signed quantity.
989 If unset, assume the remote TCP is not broken even if we do
990 not receive a window scaling option from them.
994 tcp_thin_linear_timeouts - BOOLEAN
995 Enable dynamic triggering of linear timeouts for thin streams.
996 If set, a check is performed upon retransmission by timeout to
997 determine if the stream is thin (less than 4 packets in flight).
998 As long as the stream is found to be thin, up to 6 linear
999 timeouts may be performed before exponential backoff mode is
1000 initiated. This improves retransmission latency for
1001 non-aggressive thin streams, often found to be time-dependent.
1002 For more information on thin streams, see
1003 Documentation/networking/tcp-thin.rst
1007 tcp_limit_output_bytes - INTEGER
1008 Controls TCP Small Queue limit per tcp socket.
1009 TCP bulk sender tends to increase packets in flight until it
1010 gets losses notifications. With SNDBUF autotuning, this can
1011 result in a large amount of packets queued on the local machine
1012 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
1013 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
1014 limits the number of bytes on qdisc or device to reduce artificial
1015 RTT/cwnd and reduce bufferbloat.
1017 Default: 1048576 (16 * 65536)
1019 tcp_challenge_ack_limit - INTEGER
1020 Limits number of Challenge ACK sent per second, as recommended
1021 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
1027 udp_l3mdev_accept - BOOLEAN
1028 Enabling this option allows a "global" bound socket to work
1029 across L3 master domains (e.g., VRFs) with packets capable of
1030 being received regardless of the L3 domain in which they
1031 originated. Only valid when the kernel was compiled with
1032 CONFIG_NET_L3_MASTER_DEV.
1034 Default: 0 (disabled)
1036 udp_mem - vector of 3 INTEGERs: min, pressure, max
1037 Number of pages allowed for queueing by all UDP sockets.
1039 min: Number of pages allowed for queueing by all UDP sockets.
1041 pressure: This value was introduced to follow format of tcp_mem.
1043 max: This value was introduced to follow format of tcp_mem.
1045 Default is calculated at boot time from amount of available memory.
1047 udp_rmem_min - INTEGER
1048 Minimal size of receive buffer used by UDP sockets in moderation.
1049 Each UDP socket is able to use the size for receiving data, even if
1050 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1054 udp_wmem_min - INTEGER
1055 UDP does not have tx memory accounting and this tunable has no effect.
1060 raw_l3mdev_accept - BOOLEAN
1061 Enabling this option allows a "global" bound socket to work
1062 across L3 master domains (e.g., VRFs) with packets capable of
1063 being received regardless of the L3 domain in which they
1064 originated. Only valid when the kernel was compiled with
1065 CONFIG_NET_L3_MASTER_DEV.
1067 Default: 1 (enabled)
1072 cipso_cache_enable - BOOLEAN
1073 If set, enable additions to and lookups from the CIPSO label mapping
1074 cache. If unset, additions are ignored and lookups always result in a
1075 miss. However, regardless of the setting the cache is still
1076 invalidated when required when means you can safely toggle this on and
1077 off and the cache will always be "safe".
1081 cipso_cache_bucket_size - INTEGER
1082 The CIPSO label cache consists of a fixed size hash table with each
1083 hash bucket containing a number of cache entries. This variable limits
1084 the number of entries in each hash bucket; the larger the value is, the
1085 more CIPSO label mappings that can be cached. When the number of
1086 entries in a given hash bucket reaches this limit adding new entries
1087 causes the oldest entry in the bucket to be removed to make room.
1091 cipso_rbm_optfmt - BOOLEAN
1092 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1093 the CIPSO draft specification (see Documentation/netlabel for details).
1094 This means that when set the CIPSO tag will be padded with empty
1095 categories in order to make the packet data 32-bit aligned.
1099 cipso_rbm_structvalid - BOOLEAN
1100 If set, do a very strict check of the CIPSO option when
1101 ip_options_compile() is called. If unset, relax the checks done during
1102 ip_options_compile(). Either way is "safe" as errors are caught else
1103 where in the CIPSO processing code but setting this to 0 (False) should
1104 result in less work (i.e. it should be faster) but could cause problems
1105 with other implementations that require strict checking.
1112 ip_local_port_range - 2 INTEGERS
1113 Defines the local port range that is used by TCP and UDP to
1114 choose the local port. The first number is the first, the
1115 second the last local port number.
1116 If possible, it is better these numbers have different parity
1117 (one even and one odd value).
1118 Must be greater than or equal to ip_unprivileged_port_start.
1119 The default values are 32768 and 60999 respectively.
1121 ip_local_reserved_ports - list of comma separated ranges
1122 Specify the ports which are reserved for known third-party
1123 applications. These ports will not be used by automatic port
1124 assignments (e.g. when calling connect() or bind() with port
1125 number 0). Explicit port allocation behavior is unchanged.
1127 The format used for both input and output is a comma separated
1128 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1129 10). Writing to the file will clear all previously reserved
1130 ports and update the current list with the one given in the
1133 Note that ip_local_port_range and ip_local_reserved_ports
1134 settings are independent and both are considered by the kernel
1135 when determining which ports are available for automatic port
1138 You can reserve ports which are not in the current
1139 ip_local_port_range, e.g.::
1141 $ cat /proc/sys/net/ipv4/ip_local_port_range
1143 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1146 although this is redundant. However such a setting is useful
1147 if later the port range is changed to a value that will
1148 include the reserved ports. Also keep in mind, that overlapping
1149 of these ranges may affect probability of selecting ephemeral
1150 ports which are right after block of reserved ports.
1154 ip_unprivileged_port_start - INTEGER
1155 This is a per-namespace sysctl. It defines the first
1156 unprivileged port in the network namespace. Privileged ports
1157 require root or CAP_NET_BIND_SERVICE in order to bind to them.
1158 To disable all privileged ports, set this to 0. They must not
1159 overlap with the ip_local_port_range.
1163 ip_nonlocal_bind - BOOLEAN
1164 If set, allows processes to bind() to non-local IP addresses,
1165 which can be quite useful - but may break some applications.
1169 ip_autobind_reuse - BOOLEAN
1170 By default, bind() does not select the ports automatically even if
1171 the new socket and all sockets bound to the port have SO_REUSEADDR.
1172 ip_autobind_reuse allows bind() to reuse the port and this is useful
1173 when you use bind()+connect(), but may break some applications.
1174 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1175 option should only be set by experts.
1178 ip_dynaddr - INTEGER
1179 If set non-zero, enables support for dynamic addresses.
1180 If set to a non-zero value larger than 1, a kernel log
1181 message will be printed when dynamic address rewriting
1186 ip_early_demux - BOOLEAN
1187 Optimize input packet processing down to one demux for
1188 certain kinds of local sockets. Currently we only do this
1189 for established TCP and connected UDP sockets.
1191 It may add an additional cost for pure routing workloads that
1192 reduces overall throughput, in such case you should disable it.
1196 ping_group_range - 2 INTEGERS
1197 Restrict ICMP_PROTO datagram sockets to users in the group range.
1198 The default is "1 0", meaning, that nobody (not even root) may
1199 create ping sockets. Setting it to "100 100" would grant permissions
1200 to the single group. "0 4294967295" would enable it for the world, "100
1201 4294967295" would enable it for the users, but not daemons.
1203 tcp_early_demux - BOOLEAN
1204 Enable early demux for established TCP sockets.
1208 udp_early_demux - BOOLEAN
1209 Enable early demux for connected UDP sockets. Disable this if
1210 your system could experience more unconnected load.
1214 icmp_echo_ignore_all - BOOLEAN
1215 If set non-zero, then the kernel will ignore all ICMP ECHO
1216 requests sent to it.
1220 icmp_echo_enable_probe - BOOLEAN
1221 If set to one, then the kernel will respond to RFC 8335 PROBE
1222 requests sent to it.
1226 icmp_echo_ignore_broadcasts - BOOLEAN
1227 If set non-zero, then the kernel will ignore all ICMP ECHO and
1228 TIMESTAMP requests sent to it via broadcast/multicast.
1232 icmp_ratelimit - INTEGER
1233 Limit the maximal rates for sending ICMP packets whose type matches
1234 icmp_ratemask (see below) to specific targets.
1235 0 to disable any limiting,
1236 otherwise the minimal space between responses in milliseconds.
1237 Note that another sysctl, icmp_msgs_per_sec limits the number
1238 of ICMP packets sent on all targets.
1242 icmp_msgs_per_sec - INTEGER
1243 Limit maximal number of ICMP packets sent per second from this host.
1244 Only messages whose type matches icmp_ratemask (see below) are
1245 controlled by this limit. For security reasons, the precise count
1246 of messages per second is randomized.
1250 icmp_msgs_burst - INTEGER
1251 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1252 while icmp_msgs_burst controls the burst size of these packets.
1253 For security reasons, the precise burst size is randomized.
1257 icmp_ratemask - INTEGER
1258 Mask made of ICMP types for which rates are being limited.
1260 Significant bits: IHGFEDCBA9876543210
1262 Default mask: 0000001100000011000 (6168)
1264 Bit definitions (see include/linux/icmp.h):
1266 = =========================
1268 3 Destination Unreachable [1]_
1269 4 Source Quench [1]_
1272 B Time Exceeded [1]_
1273 C Parameter Problem [1]_
1278 H Address Mask Request
1279 I Address Mask Reply
1280 = =========================
1282 .. [1] These are rate limited by default (see default mask above)
1284 icmp_ignore_bogus_error_responses - BOOLEAN
1285 Some routers violate RFC1122 by sending bogus responses to broadcast
1286 frames. Such violations are normally logged via a kernel warning.
1287 If this is set to TRUE, the kernel will not give such warnings, which
1288 will avoid log file clutter.
1292 icmp_errors_use_inbound_ifaddr - BOOLEAN
1294 If zero, icmp error messages are sent with the primary address of
1295 the exiting interface.
1297 If non-zero, the message will be sent with the primary address of
1298 the interface that received the packet that caused the icmp error.
1299 This is the behaviour many network administrators will expect from
1300 a router. And it can make debugging complicated network layouts
1303 Note that if no primary address exists for the interface selected,
1304 then the primary address of the first non-loopback interface that
1305 has one will be used regardless of this setting.
1309 igmp_max_memberships - INTEGER
1310 Change the maximum number of multicast groups we can subscribe to.
1313 Theoretical maximum value is bounded by having to send a membership
1314 report in a single datagram (i.e. the report can't span multiple
1315 datagrams, or risk confusing the switch and leaving groups you don't
1318 The number of supported groups 'M' is bounded by the number of group
1319 report entries you can fit into a single datagram of 65535 bytes.
1321 M = 65536-sizeof (ip header)/(sizeof(Group record))
1323 Group records are variable length, with a minimum of 12 bytes.
1324 So net.ipv4.igmp_max_memberships should not be set higher than:
1326 (65536-24) / 12 = 5459
1328 The value 5459 assumes no IP header options, so in practice
1329 this number may be lower.
1331 igmp_max_msf - INTEGER
1332 Maximum number of addresses allowed in the source filter list for a
1338 Controls the IGMP query robustness variable (see RFC2236 8.1).
1340 Default: 2 (as specified by RFC2236 8.1)
1342 Minimum: 1 (as specified by RFC6636 4.5)
1344 force_igmp_version - INTEGER
1345 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1346 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1347 Present timer expires.
1348 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1349 receive IGMPv2/v3 query.
1350 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1351 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1352 - 3 - Enforce to use IGMP version 3. The same react with default 0.
1356 this is not the same with force_mld_version because IGMPv3 RFC3376
1357 Security Considerations does not have clear description that we could
1358 ignore other version messages completely as MLDv2 RFC3810. So make
1359 this value as default 0 is recommended.
1361 ``conf/interface/*``
1362 changes special settings per interface (where
1363 interface" is the name of your network interface)
1366 is special, changes the settings for all interfaces
1368 log_martians - BOOLEAN
1369 Log packets with impossible addresses to kernel log.
1370 log_martians for the interface will be enabled if at least one of
1371 conf/{all,interface}/log_martians is set to TRUE,
1372 it will be disabled otherwise
1374 accept_redirects - BOOLEAN
1375 Accept ICMP redirect messages.
1376 accept_redirects for the interface will be enabled if:
1378 - both conf/{all,interface}/accept_redirects are TRUE in the case
1379 forwarding for the interface is enabled
1383 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1384 case forwarding for the interface is disabled
1386 accept_redirects for the interface will be disabled otherwise
1393 forwarding - BOOLEAN
1394 Enable IP forwarding on this interface. This controls whether packets
1395 received _on_ this interface can be forwarded.
1397 mc_forwarding - BOOLEAN
1398 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1399 and a multicast routing daemon is required.
1400 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1401 routing for the interface
1404 Integer value used to differentiate the devices by the medium they
1405 are attached to. Two devices can have different id values when
1406 the broadcast packets are received only on one of them.
1407 The default value 0 means that the device is the only interface
1408 to its medium, value of -1 means that medium is not known.
1410 Currently, it is used to change the proxy_arp behavior:
1411 the proxy_arp feature is enabled for packets forwarded between
1412 two devices attached to different media.
1417 proxy_arp for the interface will be enabled if at least one of
1418 conf/{all,interface}/proxy_arp is set to TRUE,
1419 it will be disabled otherwise
1421 proxy_arp_pvlan - BOOLEAN
1422 Private VLAN proxy arp.
1424 Basically allow proxy arp replies back to the same interface
1425 (from which the ARP request/solicitation was received).
1427 This is done to support (ethernet) switch features, like RFC
1428 3069, where the individual ports are NOT allowed to
1429 communicate with each other, but they are allowed to talk to
1430 the upstream router. As described in RFC 3069, it is possible
1431 to allow these hosts to communicate through the upstream
1432 router by proxy_arp'ing. Don't need to be used together with
1435 This technology is known by different names:
1437 In RFC 3069 it is called VLAN Aggregation.
1438 Cisco and Allied Telesyn call it Private VLAN.
1439 Hewlett-Packard call it Source-Port filtering or port-isolation.
1440 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1442 shared_media - BOOLEAN
1443 Send(router) or accept(host) RFC1620 shared media redirects.
1444 Overrides secure_redirects.
1446 shared_media for the interface will be enabled if at least one of
1447 conf/{all,interface}/shared_media is set to TRUE,
1448 it will be disabled otherwise
1452 secure_redirects - BOOLEAN
1453 Accept ICMP redirect messages only to gateways listed in the
1454 interface's current gateway list. Even if disabled, RFC1122 redirect
1457 Overridden by shared_media.
1459 secure_redirects for the interface will be enabled if at least one of
1460 conf/{all,interface}/secure_redirects is set to TRUE,
1461 it will be disabled otherwise
1465 send_redirects - BOOLEAN
1466 Send redirects, if router.
1468 send_redirects for the interface will be enabled if at least one of
1469 conf/{all,interface}/send_redirects is set to TRUE,
1470 it will be disabled otherwise
1474 bootp_relay - BOOLEAN
1475 Accept packets with source address 0.b.c.d destined
1476 not to this host as local ones. It is supposed, that
1477 BOOTP relay daemon will catch and forward such packets.
1478 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1483 Not Implemented Yet.
1485 accept_source_route - BOOLEAN
1486 Accept packets with SRR option.
1487 conf/all/accept_source_route must also be set to TRUE to accept packets
1488 with SRR option on the interface
1495 accept_local - BOOLEAN
1496 Accept packets with local source addresses. In combination with
1497 suitable routing, this can be used to direct packets between two
1498 local interfaces over the wire and have them accepted properly.
1501 route_localnet - BOOLEAN
1502 Do not consider loopback addresses as martian source or destination
1503 while routing. This enables the use of 127/8 for local routing purposes.
1508 - 0 - No source validation.
1509 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1510 Each incoming packet is tested against the FIB and if the interface
1511 is not the best reverse path the packet check will fail.
1512 By default failed packets are discarded.
1513 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1514 Each incoming packet's source address is also tested against the FIB
1515 and if the source address is not reachable via any interface
1516 the packet check will fail.
1518 Current recommended practice in RFC3704 is to enable strict mode
1519 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1520 or other complicated routing, then loose mode is recommended.
1522 The max value from conf/{all,interface}/rp_filter is used
1523 when doing source validation on the {interface}.
1525 Default value is 0. Note that some distributions enable it
1528 src_valid_mark - BOOLEAN
1529 - 0 - The fwmark of the packet is not included in reverse path
1530 route lookup. This allows for asymmetric routing configurations
1531 utilizing the fwmark in only one direction, e.g., transparent
1534 - 1 - The fwmark of the packet is included in reverse path route
1535 lookup. This permits rp_filter to function when the fwmark is
1536 used for routing traffic in both directions.
1538 This setting also affects the utilization of fmwark when
1539 performing source address selection for ICMP replies, or
1540 determining addresses stored for the IPOPT_TS_TSANDADDR and
1541 IPOPT_RR IP options.
1543 The max value from conf/{all,interface}/src_valid_mark is used.
1547 arp_filter - BOOLEAN
1548 - 1 - Allows you to have multiple network interfaces on the same
1549 subnet, and have the ARPs for each interface be answered
1550 based on whether or not the kernel would route a packet from
1551 the ARP'd IP out that interface (therefore you must use source
1552 based routing for this to work). In other words it allows control
1553 of which cards (usually 1) will respond to an arp request.
1555 - 0 - (default) The kernel can respond to arp requests with addresses
1556 from other interfaces. This may seem wrong but it usually makes
1557 sense, because it increases the chance of successful communication.
1558 IP addresses are owned by the complete host on Linux, not by
1559 particular interfaces. Only for more complex setups like load-
1560 balancing, does this behaviour cause problems.
1562 arp_filter for the interface will be enabled if at least one of
1563 conf/{all,interface}/arp_filter is set to TRUE,
1564 it will be disabled otherwise
1566 arp_announce - INTEGER
1567 Define different restriction levels for announcing the local
1568 source IP address from IP packets in ARP requests sent on
1571 - 0 - (default) Use any local address, configured on any interface
1572 - 1 - Try to avoid local addresses that are not in the target's
1573 subnet for this interface. This mode is useful when target
1574 hosts reachable via this interface require the source IP
1575 address in ARP requests to be part of their logical network
1576 configured on the receiving interface. When we generate the
1577 request we will check all our subnets that include the
1578 target IP and will preserve the source address if it is from
1579 such subnet. If there is no such subnet we select source
1580 address according to the rules for level 2.
1581 - 2 - Always use the best local address for this target.
1582 In this mode we ignore the source address in the IP packet
1583 and try to select local address that we prefer for talks with
1584 the target host. Such local address is selected by looking
1585 for primary IP addresses on all our subnets on the outgoing
1586 interface that include the target IP address. If no suitable
1587 local address is found we select the first local address
1588 we have on the outgoing interface or on all other interfaces,
1589 with the hope we will receive reply for our request and
1590 even sometimes no matter the source IP address we announce.
1592 The max value from conf/{all,interface}/arp_announce is used.
1594 Increasing the restriction level gives more chance for
1595 receiving answer from the resolved target while decreasing
1596 the level announces more valid sender's information.
1598 arp_ignore - INTEGER
1599 Define different modes for sending replies in response to
1600 received ARP requests that resolve local target IP addresses:
1602 - 0 - (default): reply for any local target IP address, configured
1604 - 1 - reply only if the target IP address is local address
1605 configured on the incoming interface
1606 - 2 - reply only if the target IP address is local address
1607 configured on the incoming interface and both with the
1608 sender's IP address are part from same subnet on this interface
1609 - 3 - do not reply for local addresses configured with scope host,
1610 only resolutions for global and link addresses are replied
1612 - 8 - do not reply for all local addresses
1614 The max value from conf/{all,interface}/arp_ignore is used
1615 when ARP request is received on the {interface}
1617 arp_notify - BOOLEAN
1618 Define mode for notification of address and device changes.
1620 == ==========================================================
1621 0 (default): do nothing
1622 1 Generate gratuitous arp requests when device is brought up
1623 or hardware address changes.
1624 == ==========================================================
1626 arp_accept - BOOLEAN
1627 Define behavior for gratuitous ARP frames who's IP is not
1628 already present in the ARP table:
1630 - 0 - don't create new entries in the ARP table
1631 - 1 - create new entries in the ARP table
1633 Both replies and requests type gratuitous arp will trigger the
1634 ARP table to be updated, if this setting is on.
1636 If the ARP table already contains the IP address of the
1637 gratuitous arp frame, the arp table will be updated regardless
1638 if this setting is on or off.
1640 arp_evict_nocarrier - BOOLEAN
1641 Clears the ARP cache on NOCARRIER events. This option is important for
1642 wireless devices where the ARP cache should not be cleared when roaming
1643 between access points on the same network. In most cases this should
1644 remain as the default (1).
1646 - 1 - (default): Clear the ARP cache on NOCARRIER events
1647 - 0 - Do not clear ARP cache on NOCARRIER events
1649 mcast_solicit - INTEGER
1650 The maximum number of multicast probes in INCOMPLETE state,
1651 when the associated hardware address is unknown. Defaults
1654 ucast_solicit - INTEGER
1655 The maximum number of unicast probes in PROBE state, when
1656 the hardware address is being reconfirmed. Defaults to 3.
1658 app_solicit - INTEGER
1659 The maximum number of probes to send to the user space ARP daemon
1660 via netlink before dropping back to multicast probes (see
1661 mcast_resolicit). Defaults to 0.
1663 mcast_resolicit - INTEGER
1664 The maximum number of multicast probes after unicast and
1665 app probes in PROBE state. Defaults to 0.
1667 disable_policy - BOOLEAN
1668 Disable IPSEC policy (SPD) for this interface
1670 disable_xfrm - BOOLEAN
1671 Disable IPSEC encryption on this interface, whatever the policy
1673 igmpv2_unsolicited_report_interval - INTEGER
1674 The interval in milliseconds in which the next unsolicited
1675 IGMPv1 or IGMPv2 report retransmit will take place.
1677 Default: 10000 (10 seconds)
1679 igmpv3_unsolicited_report_interval - INTEGER
1680 The interval in milliseconds in which the next unsolicited
1681 IGMPv3 report retransmit will take place.
1683 Default: 1000 (1 seconds)
1685 ignore_routes_with_linkdown - BOOLEAN
1686 Ignore routes whose link is down when performing a FIB lookup.
1688 promote_secondaries - BOOLEAN
1689 When a primary IP address is removed from this interface
1690 promote a corresponding secondary IP address instead of
1691 removing all the corresponding secondary IP addresses.
1693 drop_unicast_in_l2_multicast - BOOLEAN
1694 Drop any unicast IP packets that are received in link-layer
1695 multicast (or broadcast) frames.
1697 This behavior (for multicast) is actually a SHOULD in RFC
1698 1122, but is disabled by default for compatibility reasons.
1702 drop_gratuitous_arp - BOOLEAN
1703 Drop all gratuitous ARP frames, for example if there's a known
1704 good ARP proxy on the network and such frames need not be used
1705 (or in the case of 802.11, must not be used to prevent attacks.)
1711 Allows you to write a number, which can be used as required.
1715 xfrm4_gc_thresh - INTEGER
1716 (Obsolete since linux-4.14)
1717 The threshold at which we will start garbage collecting for IPv4
1718 destination cache entries. At twice this value the system will
1719 refuse new allocations.
1721 igmp_link_local_mcast_reports - BOOLEAN
1722 Enable IGMP reports for link local multicast groups in the
1728 kuznet@ms2.inr.ac.ru
1735 delon.nicolas@wanadoo.fr
1740 /proc/sys/net/ipv6/* Variables
1741 ==============================
1743 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1744 apply to IPv6 [XXX?].
1746 bindv6only - BOOLEAN
1747 Default value for IPV6_V6ONLY socket option,
1748 which restricts use of the IPv6 socket to IPv6 communication
1751 - TRUE: disable IPv4-mapped address feature
1752 - FALSE: enable IPv4-mapped address feature
1754 Default: FALSE (as specified in RFC3493)
1756 flowlabel_consistency - BOOLEAN
1757 Protect the consistency (and unicity) of flow label.
1758 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1766 auto_flowlabels - INTEGER
1767 Automatically generate flow labels based on a flow hash of the
1768 packet. This allows intermediate devices, such as routers, to
1769 identify packet flows for mechanisms like Equal Cost Multipath
1770 Routing (see RFC 6438).
1772 = ===========================================================
1773 0 automatic flow labels are completely disabled
1774 1 automatic flow labels are enabled by default, they can be
1775 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1777 2 automatic flow labels are allowed, they may be enabled on a
1778 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1779 3 automatic flow labels are enabled and enforced, they cannot
1780 be disabled by the socket option
1781 = ===========================================================
1785 flowlabel_state_ranges - BOOLEAN
1786 Split the flow label number space into two ranges. 0-0x7FFFF is
1787 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1788 is reserved for stateless flow labels as described in RFC6437.
1795 flowlabel_reflect - INTEGER
1796 Control flow label reflection. Needed for Path MTU
1797 Discovery to work with Equal Cost Multipath Routing in anycast
1798 environments. See RFC 7690 and:
1799 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1803 - 1: enabled for established flows
1805 Note that this prevents automatic flowlabel changes, as done
1806 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
1807 and "tcp: Change txhash on every SYN and RTO retransmit"
1809 - 2: enabled for TCP RESET packets (no active listener)
1810 If set, a RST packet sent in response to a SYN packet on a closed
1811 port will reflect the incoming flow label.
1813 - 4: enabled for ICMPv6 echo reply messages.
1817 fib_multipath_hash_policy - INTEGER
1818 Controls which hash policy to use for multipath routes.
1820 Default: 0 (Layer 3)
1824 - 0 - Layer 3 (source and destination addresses plus flow label)
1825 - 1 - Layer 4 (standard 5-tuple)
1826 - 2 - Layer 3 or inner Layer 3 if present
1827 - 3 - Custom multipath hash. Fields used for multipath hash calculation
1828 are determined by fib_multipath_hash_fields sysctl
1830 fib_multipath_hash_fields - UNSIGNED INTEGER
1831 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
1832 fields used for multipath hash calculation are determined by this
1835 This value is a bitmask which enables various fields for multipath hash
1838 Possible fields are:
1840 ====== ============================
1841 0x0001 Source IP address
1842 0x0002 Destination IP address
1846 0x0020 Destination port
1847 0x0040 Inner source IP address
1848 0x0080 Inner destination IP address
1849 0x0100 Inner IP protocol
1850 0x0200 Inner Flow Label
1851 0x0400 Inner source port
1852 0x0800 Inner destination port
1853 ====== ============================
1855 Default: 0x0007 (source IP, destination IP and IP protocol)
1857 anycast_src_echo_reply - BOOLEAN
1858 Controls the use of anycast addresses as source addresses for ICMPv6
1866 idgen_delay - INTEGER
1867 Controls the delay in seconds after which time to retry
1868 privacy stable address generation if a DAD conflict is
1871 Default: 1 (as specified in RFC7217)
1873 idgen_retries - INTEGER
1874 Controls the number of retries to generate a stable privacy
1875 address if a DAD conflict is detected.
1877 Default: 3 (as specified in RFC7217)
1880 Controls the MLD query robustness variable (see RFC3810 9.1).
1882 Default: 2 (as specified by RFC3810 9.1)
1884 Minimum: 1 (as specified by RFC6636 4.5)
1886 max_dst_opts_number - INTEGER
1887 Maximum number of non-padding TLVs allowed in a Destination
1888 options extension header. If this value is less than zero
1889 then unknown options are disallowed and the number of known
1890 TLVs allowed is the absolute value of this number.
1894 max_hbh_opts_number - INTEGER
1895 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1896 options extension header. If this value is less than zero
1897 then unknown options are disallowed and the number of known
1898 TLVs allowed is the absolute value of this number.
1902 max_dst_opts_length - INTEGER
1903 Maximum length allowed for a Destination options extension
1906 Default: INT_MAX (unlimited)
1908 max_hbh_length - INTEGER
1909 Maximum length allowed for a Hop-by-Hop options extension
1912 Default: INT_MAX (unlimited)
1914 skip_notify_on_dev_down - BOOLEAN
1915 Controls whether an RTM_DELROUTE message is generated for routes
1916 removed when a device is taken down or deleted. IPv4 does not
1917 generate this message; IPv6 does by default. Setting this sysctl
1918 to true skips the message, making IPv4 and IPv6 on par in relying
1919 on userspace caches to track link events and evict routes.
1921 Default: false (generate message)
1923 nexthop_compat_mode - BOOLEAN
1924 New nexthop API provides a means for managing nexthops independent of
1925 prefixes. Backwards compatibilty with old route format is enabled by
1926 default which means route dumps and notifications contain the new
1927 nexthop attribute but also the full, expanded nexthop definition.
1928 Further, updates or deletes of a nexthop configuration generate route
1929 notifications for each fib entry using the nexthop. Once a system
1930 understands the new API, this sysctl can be disabled to achieve full
1931 performance benefits of the new API by disabling the nexthop expansion
1932 and extraneous notifications.
1933 Default: true (backward compat mode)
1935 fib_notify_on_flag_change - INTEGER
1936 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
1937 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
1939 After installing a route to the kernel, user space receives an
1940 acknowledgment, which means the route was installed in the kernel,
1941 but not necessarily in hardware.
1942 It is also possible for a route already installed in hardware to change
1943 its action and therefore its flags. For example, a host route that is
1944 trapping packets can be "promoted" to perform decapsulation following
1945 the installation of an IPinIP/VXLAN tunnel.
1946 The notifications will indicate to user-space the state of the route.
1948 Default: 0 (Do not emit notifications.)
1952 - 0 - Do not emit notifications.
1953 - 1 - Emit notifications.
1954 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
1957 Define the IOAM id of this node. Uses only 24 bits out of 32 in total.
1964 ioam6_id_wide - LONG INTEGER
1965 Define the wide IOAM id of this node. Uses only 56 bits out of 64 in
1966 total. Can be different from ioam6_id.
1969 Max: 0xFFFFFFFFFFFFFF
1971 Default: 0xFFFFFFFFFFFFFF
1975 ip6frag_high_thresh - INTEGER
1976 Maximum memory used to reassemble IPv6 fragments. When
1977 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1978 the fragment handler will toss packets until ip6frag_low_thresh
1981 ip6frag_low_thresh - INTEGER
1982 See ip6frag_high_thresh
1984 ip6frag_time - INTEGER
1985 Time in seconds to keep an IPv6 fragment in memory.
1988 Change the interface-specific default settings.
1990 These settings would be used during creating new interfaces.
1994 Change all the interface-specific settings.
1996 [XXX: Other special features than forwarding?]
1998 conf/all/disable_ipv6 - BOOLEAN
1999 Changing this value is same as changing ``conf/default/disable_ipv6``
2000 setting and also all per-interface ``disable_ipv6`` settings to the same
2003 Reading this value does not have any particular meaning. It does not say
2004 whether IPv6 support is enabled or disabled. Returned value can be 1
2005 also in the case when some interface has ``disable_ipv6`` set to 0 and
2006 has configured IPv6 addresses.
2008 conf/all/forwarding - BOOLEAN
2009 Enable global IPv6 forwarding between all interfaces.
2011 IPv4 and IPv6 work differently here; e.g. netfilter must be used
2012 to control which interfaces may forward packets and which not.
2014 This also sets all interfaces' Host/Router setting
2015 'forwarding' to the specified value. See below for details.
2017 This referred to as global forwarding.
2022 fwmark_reflect - BOOLEAN
2023 Controls the fwmark of kernel-generated IPv6 reply packets that are not
2024 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
2025 If unset, these packets have a fwmark of zero. If set, they have the
2026 fwmark of the packet they are replying to.
2030 ``conf/interface/*``:
2031 Change special settings per interface.
2033 The functional behaviour for certain settings is different
2034 depending on whether local forwarding is enabled or not.
2037 Accept Router Advertisements; autoconfigure using them.
2039 It also determines whether or not to transmit Router
2040 Solicitations. If and only if the functional setting is to
2041 accept Router Advertisements, Router Solicitations will be
2044 Possible values are:
2046 == ===========================================================
2047 0 Do not accept Router Advertisements.
2048 1 Accept Router Advertisements if forwarding is disabled.
2049 2 Overrule forwarding behaviour. Accept Router Advertisements
2050 even if forwarding is enabled.
2051 == ===========================================================
2055 - enabled if local forwarding is disabled.
2056 - disabled if local forwarding is enabled.
2058 accept_ra_defrtr - BOOLEAN
2059 Learn default router in Router Advertisement.
2063 - enabled if accept_ra is enabled.
2064 - disabled if accept_ra is disabled.
2066 ra_defrtr_metric - UNSIGNED INTEGER
2067 Route metric for default route learned in Router Advertisement. This value
2068 will be assigned as metric for the default route learned via IPv6 Router
2069 Advertisement. Takes affect only if accept_ra_defrtr is enabled.
2074 Default: IP6_RT_PRIO_USER i.e. 1024.
2076 accept_ra_from_local - BOOLEAN
2077 Accept RA with source-address that is found on local machine
2078 if the RA is otherwise proper and able to be accepted.
2080 Default is to NOT accept these as it may be an un-intended
2085 - enabled if accept_ra_from_local is enabled
2086 on a specific interface.
2087 - disabled if accept_ra_from_local is disabled
2088 on a specific interface.
2090 accept_ra_min_hop_limit - INTEGER
2091 Minimum hop limit Information in Router Advertisement.
2093 Hop limit Information in Router Advertisement less than this
2094 variable shall be ignored.
2098 accept_ra_pinfo - BOOLEAN
2099 Learn Prefix Information in Router Advertisement.
2103 - enabled if accept_ra is enabled.
2104 - disabled if accept_ra is disabled.
2106 accept_ra_rt_info_min_plen - INTEGER
2107 Minimum prefix length of Route Information in RA.
2109 Route Information w/ prefix smaller than this variable shall
2114 * 0 if accept_ra_rtr_pref is enabled.
2115 * -1 if accept_ra_rtr_pref is disabled.
2117 accept_ra_rt_info_max_plen - INTEGER
2118 Maximum prefix length of Route Information in RA.
2120 Route Information w/ prefix larger than this variable shall
2125 * 0 if accept_ra_rtr_pref is enabled.
2126 * -1 if accept_ra_rtr_pref is disabled.
2128 accept_ra_rtr_pref - BOOLEAN
2129 Accept Router Preference in RA.
2133 - enabled if accept_ra is enabled.
2134 - disabled if accept_ra is disabled.
2136 accept_ra_mtu - BOOLEAN
2137 Apply the MTU value specified in RA option 5 (RFC4861). If
2138 disabled, the MTU specified in the RA will be ignored.
2142 - enabled if accept_ra is enabled.
2143 - disabled if accept_ra is disabled.
2145 accept_redirects - BOOLEAN
2150 - enabled if local forwarding is disabled.
2151 - disabled if local forwarding is enabled.
2153 accept_source_route - INTEGER
2154 Accept source routing (routing extension header).
2156 - >= 0: Accept only routing header type 2.
2157 - < 0: Do not accept routing header.
2162 Autoconfigure addresses using Prefix Information in Router
2167 - enabled if accept_ra_pinfo is enabled.
2168 - disabled if accept_ra_pinfo is disabled.
2170 dad_transmits - INTEGER
2171 The amount of Duplicate Address Detection probes to send.
2175 forwarding - INTEGER
2176 Configure interface-specific Host/Router behaviour.
2180 It is recommended to have the same setting on all
2181 interfaces; mixed router/host scenarios are rather uncommon.
2183 Possible values are:
2185 - 0 Forwarding disabled
2186 - 1 Forwarding enabled
2190 By default, Host behaviour is assumed. This means:
2192 1. IsRouter flag is not set in Neighbour Advertisements.
2193 2. If accept_ra is TRUE (default), transmit Router
2195 3. If accept_ra is TRUE (default), accept Router
2196 Advertisements (and do autoconfiguration).
2197 4. If accept_redirects is TRUE (default), accept Redirects.
2201 If local forwarding is enabled, Router behaviour is assumed.
2202 This means exactly the reverse from the above:
2204 1. IsRouter flag is set in Neighbour Advertisements.
2205 2. Router Solicitations are not sent unless accept_ra is 2.
2206 3. Router Advertisements are ignored unless accept_ra is 2.
2207 4. Redirects are ignored.
2209 Default: 0 (disabled) if global forwarding is disabled (default),
2210 otherwise 1 (enabled).
2213 Default Hop Limit to set.
2218 Default Maximum Transfer Unit
2220 Default: 1280 (IPv6 required minimum)
2222 ip_nonlocal_bind - BOOLEAN
2223 If set, allows processes to bind() to non-local IPv6 addresses,
2224 which can be quite useful - but may break some applications.
2228 router_probe_interval - INTEGER
2229 Minimum interval (in seconds) between Router Probing described
2234 router_solicitation_delay - INTEGER
2235 Number of seconds to wait after interface is brought up
2236 before sending Router Solicitations.
2240 router_solicitation_interval - INTEGER
2241 Number of seconds to wait between Router Solicitations.
2245 router_solicitations - INTEGER
2246 Number of Router Solicitations to send until assuming no
2247 routers are present.
2251 use_oif_addrs_only - BOOLEAN
2252 When enabled, the candidate source addresses for destinations
2253 routed via this interface are restricted to the set of addresses
2254 configured on this interface (vis. RFC 6724, section 4).
2258 use_tempaddr - INTEGER
2259 Preference for Privacy Extensions (RFC3041).
2261 * <= 0 : disable Privacy Extensions
2262 * == 1 : enable Privacy Extensions, but prefer public
2263 addresses over temporary addresses.
2264 * > 1 : enable Privacy Extensions and prefer temporary
2265 addresses over public addresses.
2269 * 0 (for most devices)
2270 * -1 (for point-to-point devices and loopback devices)
2272 temp_valid_lft - INTEGER
2273 valid lifetime (in seconds) for temporary addresses.
2275 Default: 172800 (2 days)
2277 temp_prefered_lft - INTEGER
2278 Preferred lifetime (in seconds) for temporary addresses.
2280 Default: 86400 (1 day)
2282 keep_addr_on_down - INTEGER
2283 Keep all IPv6 addresses on an interface down event. If set static
2284 global addresses with no expiration time are not flushed.
2287 * 0 : system default
2290 Default: 0 (addresses are removed)
2292 max_desync_factor - INTEGER
2293 Maximum value for DESYNC_FACTOR, which is a random value
2294 that ensures that clients don't synchronize with each
2295 other and generate new addresses at exactly the same time.
2296 value is in seconds.
2300 regen_max_retry - INTEGER
2301 Number of attempts before give up attempting to generate
2302 valid temporary addresses.
2306 max_addresses - INTEGER
2307 Maximum number of autoconfigured addresses per interface. Setting
2308 to zero disables the limitation. It is not recommended to set this
2309 value too large (or to zero) because it would be an easy way to
2310 crash the kernel by allowing too many addresses to be created.
2314 disable_ipv6 - BOOLEAN
2315 Disable IPv6 operation. If accept_dad is set to 2, this value
2316 will be dynamically set to TRUE if DAD fails for the link-local
2319 Default: FALSE (enable IPv6 operation)
2321 When this value is changed from 1 to 0 (IPv6 is being enabled),
2322 it will dynamically create a link-local address on the given
2323 interface and start Duplicate Address Detection, if necessary.
2325 When this value is changed from 0 to 1 (IPv6 is being disabled),
2326 it will dynamically delete all addresses and routes on the given
2327 interface. From now on it will not possible to add addresses/routes
2328 to the selected interface.
2330 accept_dad - INTEGER
2331 Whether to accept DAD (Duplicate Address Detection).
2333 == ==============================================================
2335 1 Enable DAD (default)
2336 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate
2337 link-local address has been found.
2338 == ==============================================================
2340 DAD operation and mode on a given interface will be selected according
2341 to the maximum value of conf/{all,interface}/accept_dad.
2343 force_tllao - BOOLEAN
2344 Enable sending the target link-layer address option even when
2345 responding to a unicast neighbor solicitation.
2349 Quoting from RFC 2461, section 4.4, Target link-layer address:
2351 "The option MUST be included for multicast solicitations in order to
2352 avoid infinite Neighbor Solicitation "recursion" when the peer node
2353 does not have a cache entry to return a Neighbor Advertisements
2354 message. When responding to unicast solicitations, the option can be
2355 omitted since the sender of the solicitation has the correct link-
2356 layer address; otherwise it would not have be able to send the unicast
2357 solicitation in the first place. However, including the link-layer
2358 address in this case adds little overhead and eliminates a potential
2359 race condition where the sender deletes the cached link-layer address
2360 prior to receiving a response to a previous solicitation."
2362 ndisc_notify - BOOLEAN
2363 Define mode for notification of address and device changes.
2365 * 0 - (default): do nothing
2366 * 1 - Generate unsolicited neighbour advertisements when device is brought
2367 up or hardware address changes.
2369 ndisc_tclass - INTEGER
2370 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
2371 Discovery (Router Solicitation, Router Advertisement, Neighbor
2372 Solicitation, Neighbor Advertisement, Redirect) messages.
2373 These 8 bits can be interpreted as 6 high order bits holding the DSCP
2374 value and 2 low order bits representing ECN (which you probably want
2379 ndisc_evict_nocarrier - BOOLEAN
2380 Clears the neighbor discovery table on NOCARRIER events. This option is
2381 important for wireless devices where the neighbor discovery cache should
2382 not be cleared when roaming between access points on the same network.
2383 In most cases this should remain as the default (1).
2385 - 1 - (default): Clear neighbor discover cache on NOCARRIER events.
2386 - 0 - Do not clear neighbor discovery cache on NOCARRIER events.
2388 mldv1_unsolicited_report_interval - INTEGER
2389 The interval in milliseconds in which the next unsolicited
2390 MLDv1 report retransmit will take place.
2392 Default: 10000 (10 seconds)
2394 mldv2_unsolicited_report_interval - INTEGER
2395 The interval in milliseconds in which the next unsolicited
2396 MLDv2 report retransmit will take place.
2398 Default: 1000 (1 second)
2400 force_mld_version - INTEGER
2401 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
2402 * 1 - Enforce to use MLD version 1
2403 * 2 - Enforce to use MLD version 2
2405 suppress_frag_ndisc - INTEGER
2406 Control RFC 6980 (Security Implications of IPv6 Fragmentation
2407 with IPv6 Neighbor Discovery) behavior:
2409 * 1 - (default) discard fragmented neighbor discovery packets
2410 * 0 - allow fragmented neighbor discovery packets
2412 optimistic_dad - BOOLEAN
2413 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
2415 * 0: disabled (default)
2418 Optimistic Duplicate Address Detection for the interface will be enabled
2419 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
2420 it will be disabled otherwise.
2422 use_optimistic - BOOLEAN
2423 If enabled, do not classify optimistic addresses as deprecated during
2424 source address selection. Preferred addresses will still be chosen
2425 before optimistic addresses, subject to other ranking in the source
2426 address selection algorithm.
2428 * 0: disabled (default)
2431 This will be enabled if at least one of
2432 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
2434 stable_secret - IPv6 address
2435 This IPv6 address will be used as a secret to generate IPv6
2436 addresses for link-local addresses and autoconfigured
2437 ones. All addresses generated after setting this secret will
2438 be stable privacy ones by default. This can be changed via the
2439 addrgenmode ip-link. conf/default/stable_secret is used as the
2440 secret for the namespace, the interface specific ones can
2441 overwrite that. Writes to conf/all/stable_secret are refused.
2443 It is recommended to generate this secret during installation
2444 of a system and keep it stable after that.
2446 By default the stable secret is unset.
2448 addr_gen_mode - INTEGER
2449 Defines how link-local and autoconf addresses are generated.
2451 = =================================================================
2452 0 generate address based on EUI64 (default)
2453 1 do no generate a link-local address, use EUI64 for addresses
2454 generated from autoconf
2455 2 generate stable privacy addresses, using the secret from
2456 stable_secret (RFC7217)
2457 3 generate stable privacy addresses, using a random secret if unset
2458 = =================================================================
2460 drop_unicast_in_l2_multicast - BOOLEAN
2461 Drop any unicast IPv6 packets that are received in link-layer
2462 multicast (or broadcast) frames.
2464 By default this is turned off.
2466 drop_unsolicited_na - BOOLEAN
2467 Drop all unsolicited neighbor advertisements, for example if there's
2468 a known good NA proxy on the network and such frames need not be used
2469 (or in the case of 802.11, must not be used to prevent attacks.)
2471 By default this is turned off.
2473 accept_untracked_na - BOOLEAN
2474 Add a new neighbour cache entry in STALE state for routers on receiving a
2475 neighbour advertisement (either solicited or unsolicited) with target
2476 link-layer address option specified if no neighbour entry is already
2477 present for the advertised IPv6 address. Without this knob, NAs received
2478 for untracked addresses (absent in neighbour cache) are silently ignored.
2480 This is as per router-side behaviour documented in RFC9131.
2482 This has lower precedence than drop_unsolicited_na.
2484 This will optimize the return path for the initial off-link communication
2485 that is initiated by a directly connected host, by ensuring that
2486 the first-hop router which turns on this setting doesn't have to
2487 buffer the initial return packets to do neighbour-solicitation.
2488 The prerequisite is that the host is configured to send
2489 unsolicited neighbour advertisements on interface bringup.
2490 This setting should be used in conjunction with the ndisc_notify setting
2491 on the host to satisfy this prerequisite.
2493 By default this is turned off.
2495 enhanced_dad - BOOLEAN
2496 Include a nonce option in the IPv6 neighbor solicitation messages used for
2497 duplicate address detection per RFC7527. A received DAD NS will only signal
2498 a duplicate address if the nonce is different. This avoids any false
2499 detection of duplicates due to loopback of the NS messages that we send.
2500 The nonce option will be sent on an interface unless both of
2501 conf/{all,interface}/enhanced_dad are set to FALSE.
2509 Limit the maximal rates for sending ICMPv6 messages.
2511 0 to disable any limiting,
2512 otherwise the minimal space between responses in milliseconds.
2516 ratemask - list of comma separated ranges
2517 For ICMPv6 message types matching the ranges in the ratemask, limit
2518 the sending of the message according to ratelimit parameter.
2520 The format used for both input and output is a comma separated
2521 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
2522 129). Writing to the file will clear all previous ranges of ICMPv6
2523 message types and update the current list with the input.
2525 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
2526 for numerical values of ICMPv6 message types, e.g. echo request is 128
2527 and echo reply is 129.
2529 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
2531 echo_ignore_all - BOOLEAN
2532 If set non-zero, then the kernel will ignore all ICMP ECHO
2533 requests sent to it over the IPv6 protocol.
2537 echo_ignore_multicast - BOOLEAN
2538 If set non-zero, then the kernel will ignore all ICMP ECHO
2539 requests sent to it over the IPv6 protocol via multicast.
2543 echo_ignore_anycast - BOOLEAN
2544 If set non-zero, then the kernel will ignore all ICMP ECHO
2545 requests sent to it over the IPv6 protocol destined to anycast address.
2549 xfrm6_gc_thresh - INTEGER
2550 (Obsolete since linux-4.14)
2551 The threshold at which we will start garbage collecting for IPv6
2552 destination cache entries. At twice this value the system will
2553 refuse new allocations.
2557 Pekka Savola <pekkas@netcore.fi>
2558 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2561 /proc/sys/net/bridge/* Variables:
2562 =================================
2564 bridge-nf-call-arptables - BOOLEAN
2565 - 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2570 bridge-nf-call-iptables - BOOLEAN
2571 - 1 : pass bridged IPv4 traffic to iptables' chains.
2576 bridge-nf-call-ip6tables - BOOLEAN
2577 - 1 : pass bridged IPv6 traffic to ip6tables' chains.
2582 bridge-nf-filter-vlan-tagged - BOOLEAN
2583 - 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2588 bridge-nf-filter-pppoe-tagged - BOOLEAN
2589 - 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2594 bridge-nf-pass-vlan-input-dev - BOOLEAN
2595 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2596 interface on the bridge and set the netfilter input device to the
2597 vlan. This allows use of e.g. "iptables -i br0.1" and makes the
2598 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no
2599 matching vlan interface is found, or this switch is off, the input
2600 device is set to the bridge interface.
2602 - 0: disable bridge netfilter vlan interface lookup.
2606 ``proc/sys/net/sctp/*`` Variables:
2607 ==================================
2609 addip_enable - BOOLEAN
2610 Enable or disable extension of Dynamic Address Reconfiguration
2611 (ADD-IP) functionality specified in RFC5061. This extension provides
2612 the ability to dynamically add and remove new addresses for the SCTP
2615 1: Enable extension.
2617 0: Disable extension.
2622 Enable or disable pf (pf is short for potentially failed) state. A value
2623 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2624 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2625 Since pf_retrans and path_max_retrans can be changed by userspace
2626 application, sometimes user expects to disable pf state by the value of
2627 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2628 or path_max_retrans is changed by the user application, this pf state is
2629 enabled. As such, it is necessary to add this to dynamically enable
2630 and disable pf state. See:
2631 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2641 Unset or enable/disable pf (pf is short for potentially failed) state
2642 exposure. Applications can control the exposure of the PF path state
2643 in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2644 sockopt. When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2645 SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2646 can be got via SCTP_GET_PEER_ADDR_INFO sockopt; When it's enabled,
2647 a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2648 SCTP_PF state and a SCTP_PF-state transport info can be got via
2649 SCTP_GET_PEER_ADDR_INFO sockopt; When it's diabled, no
2650 SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2651 trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2654 0: Unset pf state exposure, Compatible with old applications.
2656 1: Disable pf state exposure.
2658 2: Enable pf state exposure.
2662 addip_noauth_enable - BOOLEAN
2663 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2664 authentication to protect the operations of adding or removing new
2665 addresses. This requirement is mandated so that unauthorized hosts
2666 would not be able to hijack associations. However, older
2667 implementations may not have implemented this requirement while
2668 allowing the ADD-IP extension. For reasons of interoperability,
2669 we provide this variable to control the enforcement of the
2670 authentication requirement.
2672 == ===============================================================
2673 1 Allow ADD-IP extension to be used without authentication. This
2674 should only be set in a closed environment for interoperability
2675 with older implementations.
2677 0 Enforce the authentication requirement
2678 == ===============================================================
2682 auth_enable - BOOLEAN
2683 Enable or disable Authenticated Chunks extension. This extension
2684 provides the ability to send and receive authenticated chunks and is
2685 required for secure operation of Dynamic Address Reconfiguration
2688 - 1: Enable this extension.
2689 - 0: Disable this extension.
2693 prsctp_enable - BOOLEAN
2694 Enable or disable the Partial Reliability extension (RFC3758) which
2695 is used to notify peers that a given DATA should no longer be expected.
2697 - 1: Enable extension
2703 The limit of the number of new packets that can be initially sent. It
2704 controls how bursty the generated traffic can be.
2708 association_max_retrans - INTEGER
2709 Set the maximum number for retransmissions that an association can
2710 attempt deciding that the remote end is unreachable. If this value
2711 is exceeded, the association is terminated.
2715 max_init_retransmits - INTEGER
2716 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2717 that an association will attempt before declaring the destination
2718 unreachable and terminating.
2722 path_max_retrans - INTEGER
2723 The maximum number of retransmissions that will be attempted on a given
2724 path. Once this threshold is exceeded, the path is considered
2725 unreachable, and new traffic will use a different path when the
2726 association is multihomed.
2730 pf_retrans - INTEGER
2731 The number of retransmissions that will be attempted on a given path
2732 before traffic is redirected to an alternate transport (should one
2733 exist). Note this is distinct from path_max_retrans, as a path that
2734 passes the pf_retrans threshold can still be used. Its only
2735 deprioritized when a transmission path is selected by the stack. This
2736 setting is primarily used to enable fast failover mechanisms without
2737 having to reduce path_max_retrans to a very low value. See:
2738 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2739 for details. Note also that a value of pf_retrans > path_max_retrans
2740 disables this feature. Since both pf_retrans and path_max_retrans can
2741 be changed by userspace application, a variable pf_enable is used to
2746 ps_retrans - INTEGER
2747 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
2748 from section-5 "Primary Path Switchover" in rfc7829. The primary path
2749 will be changed to another active path when the path error counter on
2750 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
2751 to continue data transmission on a new working path even when the old
2752 primary destination address becomes active again". Note this feature
2753 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
2754 and its value can't be less than 'pf_retrans' when changing by sysctl.
2758 rto_initial - INTEGER
2759 The initial round trip timeout value in milliseconds that will be used
2760 in calculating round trip times. This is the initial time interval
2761 for retransmissions.
2766 The maximum value (in milliseconds) of the round trip timeout. This
2767 is the largest time interval that can elapse between retransmissions.
2772 The minimum value (in milliseconds) of the round trip timeout. This
2773 is the smallest time interval the can elapse between retransmissions.
2777 hb_interval - INTEGER
2778 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2779 are sent at the specified interval on idle paths to probe the state of
2780 a given path between 2 associations.
2784 sack_timeout - INTEGER
2785 The amount of time (in milliseconds) that the implementation will wait
2790 valid_cookie_life - INTEGER
2791 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2792 is used during association establishment.
2796 cookie_preserve_enable - BOOLEAN
2797 Enable or disable the ability to extend the lifetime of the SCTP cookie
2798 that is used during the establishment phase of SCTP association
2800 - 1: Enable cookie lifetime extension.
2805 cookie_hmac_alg - STRING
2806 Select the hmac algorithm used when generating the cookie value sent by
2807 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2814 Ability to assign md5 or sha1 as the selected alg is predicated on the
2815 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2816 CONFIG_CRYPTO_SHA1).
2818 Default: Dependent on configuration. MD5 if available, else SHA1 if
2819 available, else none.
2821 rcvbuf_policy - INTEGER
2822 Determines if the receive buffer is attributed to the socket or to
2823 association. SCTP supports the capability to create multiple
2824 associations on a single socket. When using this capability, it is
2825 possible that a single stalled association that's buffering a lot
2826 of data may block other associations from delivering their data by
2827 consuming all of the receive buffer space. To work around this,
2828 the rcvbuf_policy could be set to attribute the receiver buffer space
2829 to each association instead of the socket. This prevents the described
2832 - 1: rcvbuf space is per association
2833 - 0: rcvbuf space is per socket
2837 sndbuf_policy - INTEGER
2838 Similar to rcvbuf_policy above, this applies to send buffer space.
2840 - 1: Send buffer is tracked per association
2841 - 0: Send buffer is tracked per socket.
2845 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2846 Number of pages allowed for queueing by all SCTP sockets.
2848 min: Below this number of pages SCTP is not bothered about its
2849 memory appetite. When amount of memory allocated by SCTP exceeds
2850 this number, SCTP starts to moderate memory usage.
2852 pressure: This value was introduced to follow format of tcp_mem.
2854 max: Number of pages allowed for queueing by all SCTP sockets.
2856 Default is calculated at boot time from amount of available memory.
2858 sctp_rmem - vector of 3 INTEGERs: min, default, max
2859 Only the first value ("min") is used, "default" and "max" are
2862 min: Minimal size of receive buffer used by SCTP socket.
2863 It is guaranteed to each SCTP socket (but not association) even
2864 under moderate memory pressure.
2868 sctp_wmem - vector of 3 INTEGERs: min, default, max
2869 Currently this tunable has no effect.
2871 addr_scope_policy - INTEGER
2872 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2874 - 0 - Disable IPv4 address scoping
2875 - 1 - Enable IPv4 address scoping
2876 - 2 - Follow draft but allow IPv4 private addresses
2877 - 3 - Follow draft but allow IPv4 link local addresses
2882 The listening port for the local UDP tunneling sock. Normally it's
2883 using the IANA-assigned UDP port number 9899 (sctp-tunneling).
2885 This UDP sock is used for processing the incoming UDP-encapsulated
2886 SCTP packets (from RFC6951), and shared by all applications in the
2887 same net namespace. This UDP sock will be closed when the value is
2890 The value will also be used to set the src port of the UDP header
2891 for the outgoing UDP-encapsulated SCTP packets. For the dest port,
2892 please refer to 'encap_port' below.
2896 encap_port - INTEGER
2897 The default remote UDP encapsulation port.
2899 This value is used to set the dest port of the UDP header for the
2900 outgoing UDP-encapsulated SCTP packets by default. Users can also
2901 change the value for each sock/asoc/transport by using setsockopt.
2902 For further information, please refer to RFC6951.
2904 Note that when connecting to a remote server, the client should set
2905 this to the port that the UDP tunneling sock on the peer server is
2906 listening to and the local UDP tunneling sock on the client also
2907 must be started. On the server, it would get the encap_port from
2908 the incoming packet's source port.
2912 plpmtud_probe_interval - INTEGER
2913 The time interval (in milliseconds) for the PLPMTUD probe timer,
2914 which is configured to expire after this period to receive an
2915 acknowledgment to a probe packet. This is also the time interval
2916 between the probes for the current pmtu when the probe search
2919 PLPMTUD will be disabled when 0 is set, and other values for it
2924 reconf_enable - BOOLEAN
2925 Enable or disable extension of Stream Reconfiguration functionality
2926 specified in RFC6525. This extension provides the ability to "reset"
2927 a stream, and it includes the Parameters of "Outgoing/Incoming SSN
2928 Reset", "SSN/TSN Reset" and "Add Outgoing/Incoming Streams".
2930 - 1: Enable extension.
2931 - 0: Disable extension.
2935 intl_enable - BOOLEAN
2936 Enable or disable extension of User Message Interleaving functionality
2937 specified in RFC8260. This extension allows the interleaving of user
2938 messages sent on different streams. With this feature enabled, I-DATA
2939 chunk will replace DATA chunk to carry user messages if also supported
2940 by the peer. Note that to use this feature, one needs to set this option
2941 to 1 and also needs to set socket options SCTP_FRAGMENT_INTERLEAVE to 2
2942 and SCTP_INTERLEAVING_SUPPORTED to 1.
2944 - 1: Enable extension.
2945 - 0: Disable extension.
2949 ecn_enable - BOOLEAN
2950 Control use of Explicit Congestion Notification (ECN) by SCTP.
2951 Like in TCP, ECN is used only when both ends of the SCTP connection
2952 indicate support for it. This feature is useful in avoiding losses
2953 due to congestion by allowing supporting routers to signal congestion
2954 before having to drop packets.
2962 ``/proc/sys/net/core/*``
2963 ========================
2965 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
2968 ``/proc/sys/net/unix/*``
2969 ========================
2971 max_dgram_qlen - INTEGER
2972 The maximum length of dgram socket receive queue