1 .. SPDX-License-Identifier: GPL-2.0
7 /proc/sys/net/ipv4/* Variables
8 ==============================
11 - 0 - disabled (default)
14 Forward Packets between interfaces.
16 This variable is special, its change resets all configuration
17 parameters to their default state (RFC1122 for hosts, RFC1812
20 ip_default_ttl - INTEGER
21 Default value of TTL field (Time To Live) for outgoing (but not
22 forwarded) IP packets. Should be between 1 and 255 inclusive.
23 Default: 64 (as recommended by RFC1700)
25 ip_no_pmtu_disc - INTEGER
26 Disable Path MTU Discovery. If enabled in mode 1 and a
27 fragmentation-required ICMP is received, the PMTU to this
28 destination will be set to min_pmtu (see below). You will need
29 to raise min_pmtu to the smallest interface MTU on your system
30 manually if you want to avoid locally generated fragments.
32 In mode 2 incoming Path MTU Discovery messages will be
33 discarded. Outgoing frames are handled the same as in mode 1,
34 implicitly setting IP_PMTUDISC_DONT on every created socket.
36 Mode 3 is a hardened pmtu discover mode. The kernel will only
37 accept fragmentation-needed errors if the underlying protocol
38 can verify them besides a plain socket lookup. Current
39 protocols for which pmtu events will be honored are TCP, SCTP
40 and DCCP as they verify e.g. the sequence number or the
41 association. This mode should not be enabled globally but is
42 only intended to secure e.g. name servers in namespaces where
43 TCP path mtu must still work but path MTU information of other
44 protocols should be discarded. If enabled globally this mode
45 could break other protocols.
52 default 552 - minimum discovered Path MTU
54 ip_forward_use_pmtu - BOOLEAN
55 By default we don't trust protocol path MTUs while forwarding
56 because they could be easily forged and can lead to unwanted
57 fragmentation by the router.
58 You only need to enable this if you have user-space software
59 which tries to discover path mtus by itself and depends on the
60 kernel honoring this information. This is normally not the
70 fwmark_reflect - BOOLEAN
71 Controls the fwmark of kernel-generated IPv4 reply packets that are not
72 associated with a socket for example, TCP RSTs or ICMP echo replies).
73 If unset, these packets have a fwmark of zero. If set, they have the
74 fwmark of the packet they are replying to.
78 fib_multipath_use_neigh - BOOLEAN
79 Use status of existing neighbor entry when determining nexthop for
80 multipath routes. If disabled, neighbor information is not used and
81 packets could be directed to a failed nexthop. Only valid for kernels
82 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
91 fib_multipath_hash_policy - INTEGER
92 Controls which hash policy to use for multipath routes. Only valid
93 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
101 - 2 - Layer 3 or inner Layer 3 if present
103 fib_sync_mem - UNSIGNED INTEGER
104 Amount of dirty memory from fib entries that can be backlogged before
105 synchronize_rcu is forced.
107 Default: 512kB Minimum: 64kB Maximum: 64MB
109 ip_forward_update_priority - INTEGER
110 Whether to update SKB priority from "TOS" field in IPv4 header after it
111 is forwarded. The new SKB priority is mapped from TOS field value
112 according to an rt_tos2priority table (see e.g. man tc-prio).
114 Default: 1 (Update priority.)
118 - 0 - Do not update priority.
119 - 1 - Update priority.
121 route/max_size - INTEGER
122 Maximum number of routes allowed in the kernel. Increase
123 this when using large numbers of interfaces and/or routes.
125 From linux kernel 3.6 onwards, this is deprecated for ipv4
126 as route cache is no longer used.
128 neigh/default/gc_thresh1 - INTEGER
129 Minimum number of entries to keep. Garbage collector will not
130 purge entries if there are fewer than this number.
134 neigh/default/gc_thresh2 - INTEGER
135 Threshold when garbage collector becomes more aggressive about
136 purging entries. Entries older than 5 seconds will be cleared
137 when over this number.
141 neigh/default/gc_thresh3 - INTEGER
142 Maximum number of non-PERMANENT neighbor entries allowed. Increase
143 this when using large numbers of interfaces and when communicating
144 with large numbers of directly-connected peers.
148 neigh/default/unres_qlen_bytes - INTEGER
149 The maximum number of bytes which may be used by packets
150 queued for each unresolved address by other network layers.
153 Setting negative value is meaningless and will return error.
155 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
157 Exact value depends on architecture and kernel options,
158 but should be enough to allow queuing 256 packets
161 neigh/default/unres_qlen - INTEGER
162 The maximum number of packets which may be queued for each
163 unresolved address by other network layers.
165 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
167 Prior to linux 3.3, the default value is 3 which may cause
168 unexpected packet loss. The current default value is calculated
169 according to default value of unres_qlen_bytes and true size of
174 mtu_expires - INTEGER
175 Time, in seconds, that cached PMTU information is kept.
177 min_adv_mss - INTEGER
178 The advertised MSS depends on the first hop route MTU, but will
179 never be lower than this setting.
183 ipfrag_high_thresh - LONG INTEGER
184 Maximum memory used to reassemble IP fragments.
186 ipfrag_low_thresh - LONG INTEGER
187 (Obsolete since linux-4.17)
188 Maximum memory used to reassemble IP fragments before the kernel
189 begins to remove incomplete fragment queues to free up resources.
190 The kernel still accepts new fragments for defragmentation.
192 ipfrag_time - INTEGER
193 Time in seconds to keep an IP fragment in memory.
195 ipfrag_max_dist - INTEGER
196 ipfrag_max_dist is a non-negative integer value which defines the
197 maximum "disorder" which is allowed among fragments which share a
198 common IP source address. Note that reordering of packets is
199 not unusual, but if a large number of fragments arrive from a source
200 IP address while a particular fragment queue remains incomplete, it
201 probably indicates that one or more fragments belonging to that queue
202 have been lost. When ipfrag_max_dist is positive, an additional check
203 is done on fragments before they are added to a reassembly queue - if
204 ipfrag_max_dist (or more) fragments have arrived from a particular IP
205 address between additions to any IP fragment queue using that source
206 address, it's presumed that one or more fragments in the queue are
207 lost. The existing fragment queue will be dropped, and a new one
208 started. An ipfrag_max_dist value of zero disables this check.
210 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
211 result in unnecessarily dropping fragment queues when normal
212 reordering of packets occurs, which could lead to poor application
213 performance. Using a very large value, e.g. 50000, increases the
214 likelihood of incorrectly reassembling IP fragments that originate
215 from different IP datagrams, which could result in data corruption.
221 inet_peer_threshold - INTEGER
222 The approximate size of the storage. Starting from this threshold
223 entries will be thrown aggressively. This threshold also determines
224 entries' time-to-live and time intervals between garbage collection
225 passes. More entries, less time-to-live, less GC interval.
227 inet_peer_minttl - INTEGER
228 Minimum time-to-live of entries. Should be enough to cover fragment
229 time-to-live on the reassembling side. This minimum time-to-live is
230 guaranteed if the pool size is less than inet_peer_threshold.
233 inet_peer_maxttl - INTEGER
234 Maximum time-to-live of entries. Unused entries will expire after
235 this period of time if there is no memory pressure on the pool (i.e.
236 when the number of entries in the pool is very small).
243 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
244 Defaults to 4096. (Was 128 before linux-5.4)
245 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
247 tcp_abort_on_overflow - BOOLEAN
248 If listening service is too slow to accept new connections,
249 reset them. Default state is FALSE. It means that if overflow
250 occurred due to a burst, connection will recover. Enable this
251 option _only_ if you are really sure that listening daemon
252 cannot be tuned to accept connections faster. Enabling this
253 option can harm clients of your server.
255 tcp_adv_win_scale - INTEGER
256 Count buffering overhead as bytes/2^tcp_adv_win_scale
257 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
260 Possible values are [-31, 31], inclusive.
264 tcp_allowed_congestion_control - STRING
265 Show/set the congestion control choices available to non-privileged
266 processes. The list is a subset of those listed in
267 tcp_available_congestion_control.
269 Default is "reno" and the default setting (tcp_congestion_control).
271 tcp_app_win - INTEGER
272 Reserve max(window/2^tcp_app_win, mss) of window for application
273 buffer. Value 0 is special, it means that nothing is reserved.
275 Possible values are [0, 31], inclusive.
279 tcp_autocorking - BOOLEAN
280 Enable TCP auto corking :
281 When applications do consecutive small write()/sendmsg() system calls,
282 we try to coalesce these small writes as much as possible, to lower
283 total amount of sent packets. This is done if at least one prior
284 packet for the flow is waiting in Qdisc queues or device transmit
285 queue. Applications can still use TCP_CORK for optimal behavior
286 when they know how/when to uncork their sockets.
290 tcp_available_congestion_control - STRING
291 Shows the available congestion control choices that are registered.
292 More congestion control algorithms may be available as modules,
295 tcp_base_mss - INTEGER
296 The initial value of search_low to be used by the packetization layer
297 Path MTU discovery (MTU probing). If MTU probing is enabled,
298 this is the initial MSS used by the connection.
300 tcp_mtu_probe_floor - INTEGER
301 If MTU probing is enabled this caps the minimum MSS used for search_low
306 tcp_min_snd_mss - INTEGER
307 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
308 as described in RFC 1122 and RFC 6691.
310 If this ADVMSS option is smaller than tcp_min_snd_mss,
311 it is silently capped to tcp_min_snd_mss.
313 Default : 48 (at least 8 bytes of payload per segment)
315 tcp_congestion_control - STRING
316 Set the congestion control algorithm to be used for new
317 connections. The algorithm "reno" is always available, but
318 additional choices may be available based on kernel configuration.
319 Default is set as part of kernel configuration.
320 For passive connections, the listener congestion control choice
323 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
326 Allows TCP to send "duplicate" SACKs.
328 tcp_early_retrans - INTEGER
329 Tail loss probe (TLP) converts RTOs occurring due to tail
330 losses into fast recovery (draft-ietf-tcpm-rack). Note that
331 TLP requires RACK to function properly (see tcp_recovery below)
341 Control use of Explicit Congestion Notification (ECN) by TCP.
342 ECN is used only when both ends of the TCP connection indicate
343 support for it. This feature is useful in avoiding losses due
344 to congestion by allowing supporting routers to signal
345 congestion before having to drop packets.
349 = =====================================================
350 0 Disable ECN. Neither initiate nor accept ECN.
351 1 Enable ECN when requested by incoming connections and
352 also request ECN on outgoing connection attempts.
353 2 Enable ECN when requested by incoming connections
354 but do not request ECN on outgoing connections.
355 = =====================================================
359 tcp_ecn_fallback - BOOLEAN
360 If the kernel detects that ECN connection misbehaves, enable fall
361 back to non-ECN. Currently, this knob implements the fallback
362 from RFC3168, section 6.1.1.1., but we reserve that in future,
363 additional detection mechanisms could be implemented under this
364 knob. The value is not used, if tcp_ecn or per route (or congestion
365 control) ECN settings are disabled.
367 Default: 1 (fallback enabled)
370 This is a legacy option, it has no effect anymore.
372 tcp_fin_timeout - INTEGER
373 The length of time an orphaned (no longer referenced by any
374 application) connection will remain in the FIN_WAIT_2 state
375 before it is aborted at the local end. While a perfectly
376 valid "receive only" state for an un-orphaned connection, an
377 orphaned connection in FIN_WAIT_2 state could otherwise wait
378 forever for the remote to close its end of the connection.
385 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
386 F-RTO is an enhanced recovery algorithm for TCP retransmission
387 timeouts. It is particularly beneficial in networks where the
388 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
389 modification. It does not require any support from the peer.
391 By default it's enabled with a non-zero value. 0 disables F-RTO.
393 tcp_fwmark_accept - BOOLEAN
394 If set, incoming connections to listening sockets that do not have a
395 socket mark will set the mark of the accepting socket to the fwmark of
396 the incoming SYN packet. This will cause all packets on that connection
397 (starting from the first SYNACK) to be sent with that fwmark. The
398 listening socket's mark is unchanged. Listening sockets that already
399 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
404 tcp_invalid_ratelimit - INTEGER
405 Limit the maximal rate for sending duplicate acknowledgments
406 in response to incoming TCP packets that are for an existing
407 connection but that are invalid due to any of these reasons:
409 (a) out-of-window sequence number,
410 (b) out-of-window acknowledgment number, or
411 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
413 This can help mitigate simple "ack loop" DoS attacks, wherein
414 a buggy or malicious middlebox or man-in-the-middle can
415 rewrite TCP header fields in manner that causes each endpoint
416 to think that the other is sending invalid TCP segments, thus
417 causing each side to send an unterminating stream of duplicate
418 acknowledgments for invalid segments.
420 Using 0 disables rate-limiting of dupacks in response to
421 invalid segments; otherwise this value specifies the minimal
422 space between sending such dupacks, in milliseconds.
424 Default: 500 (milliseconds).
426 tcp_keepalive_time - INTEGER
427 How often TCP sends out keepalive messages when keepalive is enabled.
430 tcp_keepalive_probes - INTEGER
431 How many keepalive probes TCP sends out, until it decides that the
432 connection is broken. Default value: 9.
434 tcp_keepalive_intvl - INTEGER
435 How frequently the probes are send out. Multiplied by
436 tcp_keepalive_probes it is time to kill not responding connection,
437 after probes started. Default value: 75sec i.e. connection
438 will be aborted after ~11 minutes of retries.
440 tcp_l3mdev_accept - BOOLEAN
441 Enables child sockets to inherit the L3 master device index.
442 Enabling this option allows a "global" listen socket to work
443 across L3 master domains (e.g., VRFs) with connected sockets
444 derived from the listen socket to be bound to the L3 domain in
445 which the packets originated. Only valid when the kernel was
446 compiled with CONFIG_NET_L3_MASTER_DEV.
448 Default: 0 (disabled)
450 tcp_low_latency - BOOLEAN
451 This is a legacy option, it has no effect anymore.
453 tcp_max_orphans - INTEGER
454 Maximal number of TCP sockets not attached to any user file handle,
455 held by system. If this number is exceeded orphaned connections are
456 reset immediately and warning is printed. This limit exists
457 only to prevent simple DoS attacks, you _must_ not rely on this
458 or lower the limit artificially, but rather increase it
459 (probably, after increasing installed memory),
460 if network conditions require more than default value,
461 and tune network services to linger and kill such states
462 more aggressively. Let me to remind again: each orphan eats
463 up to ~64K of unswappable memory.
465 tcp_max_syn_backlog - INTEGER
466 Maximal number of remembered connection requests (SYN_RECV),
467 which have not received an acknowledgment from connecting client.
469 This is a per-listener limit.
471 The minimal value is 128 for low memory machines, and it will
472 increase in proportion to the memory of machine.
474 If server suffers from overload, try increasing this number.
476 Remember to also check /proc/sys/net/core/somaxconn
477 A SYN_RECV request socket consumes about 304 bytes of memory.
479 tcp_max_tw_buckets - INTEGER
480 Maximal number of timewait sockets held by system simultaneously.
481 If this number is exceeded time-wait socket is immediately destroyed
482 and warning is printed. This limit exists only to prevent
483 simple DoS attacks, you _must_ not lower the limit artificially,
484 but rather increase it (probably, after increasing installed memory),
485 if network conditions require more than default value.
487 tcp_mem - vector of 3 INTEGERs: min, pressure, max
488 min: below this number of pages TCP is not bothered about its
491 pressure: when amount of memory allocated by TCP exceeds this number
492 of pages, TCP moderates its memory consumption and enters memory
493 pressure mode, which is exited when memory consumption falls
496 max: number of pages allowed for queueing by all TCP sockets.
498 Defaults are calculated at boot time from amount of available
501 tcp_min_rtt_wlen - INTEGER
502 The window length of the windowed min filter to track the minimum RTT.
503 A shorter window lets a flow more quickly pick up new (higher)
504 minimum RTT when it is moved to a longer path (e.g., due to traffic
505 engineering). A longer window makes the filter more resistant to RTT
506 inflations such as transient congestion. The unit is seconds.
508 Possible values: 0 - 86400 (1 day)
512 tcp_moderate_rcvbuf - BOOLEAN
513 If set, TCP performs receive buffer auto-tuning, attempting to
514 automatically size the buffer (no greater than tcp_rmem[2]) to
515 match the size required by the path for full throughput. Enabled by
518 tcp_mtu_probing - INTEGER
519 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
523 - 1 - Disabled by default, enabled when an ICMP black hole detected
524 - 2 - Always enabled, use initial MSS of tcp_base_mss.
526 tcp_probe_interval - UNSIGNED INTEGER
527 Controls how often to start TCP Packetization-Layer Path MTU
528 Discovery reprobe. The default is reprobing every 10 minutes as
531 tcp_probe_threshold - INTEGER
532 Controls when TCP Packetization-Layer Path MTU Discovery probing
533 will stop in respect to the width of search range in bytes. Default
536 tcp_no_metrics_save - BOOLEAN
537 By default, TCP saves various connection metrics in the route cache
538 when the connection closes, so that connections established in the
539 near future can use these to set initial conditions. Usually, this
540 increases overall performance, but may sometimes cause performance
541 degradation. If set, TCP will not cache metrics on closing
544 tcp_no_ssthresh_metrics_save - BOOLEAN
545 Controls whether TCP saves ssthresh metrics in the route cache.
547 Default is 1, which disables ssthresh metrics.
549 tcp_orphan_retries - INTEGER
550 This value influences the timeout of a locally closed TCP connection,
551 when RTO retransmissions remain unacknowledged.
552 See tcp_retries2 for more details.
554 The default value is 8.
556 If your machine is a loaded WEB server,
557 you should think about lowering this value, such sockets
558 may consume significant resources. Cf. tcp_max_orphans.
560 tcp_recovery - INTEGER
561 This value is a bitmap to enable various experimental loss recovery
564 ========= =============================================================
565 RACK: 0x1 enables the RACK loss detection for fast detection of lost
566 retransmissions and tail drops. It also subsumes and disables
567 RFC6675 recovery for SACK connections.
569 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
571 RACK: 0x4 disables RACK's DUPACK threshold heuristic
572 ========= =============================================================
576 tcp_reordering - INTEGER
577 Initial reordering level of packets in a TCP stream.
578 TCP stack can then dynamically adjust flow reordering level
579 between this initial value and tcp_max_reordering
583 tcp_max_reordering - INTEGER
584 Maximal reordering level of packets in a TCP stream.
585 300 is a fairly conservative value, but you might increase it
586 if paths are using per packet load balancing (like bonding rr mode)
590 tcp_retrans_collapse - BOOLEAN
591 Bug-to-bug compatibility with some broken printers.
592 On retransmit try to send bigger packets to work around bugs in
595 tcp_retries1 - INTEGER
596 This value influences the time, after which TCP decides, that
597 something is wrong due to unacknowledged RTO retransmissions,
598 and reports this suspicion to the network layer.
599 See tcp_retries2 for more details.
601 RFC 1122 recommends at least 3 retransmissions, which is the
604 tcp_retries2 - INTEGER
605 This value influences the timeout of an alive TCP connection,
606 when RTO retransmissions remain unacknowledged.
607 Given a value of N, a hypothetical TCP connection following
608 exponential backoff with an initial RTO of TCP_RTO_MIN would
609 retransmit N times before killing the connection at the (N+1)th RTO.
611 The default value of 15 yields a hypothetical timeout of 924.6
612 seconds and is a lower bound for the effective timeout.
613 TCP will effectively time out at the first RTO which exceeds the
614 hypothetical timeout.
616 RFC 1122 recommends at least 100 seconds for the timeout,
617 which corresponds to a value of at least 8.
619 tcp_rfc1337 - BOOLEAN
620 If set, the TCP stack behaves conforming to RFC1337. If unset,
621 we are not conforming to RFC, but prevent TCP TIME_WAIT
626 tcp_rmem - vector of 3 INTEGERs: min, default, max
627 min: Minimal size of receive buffer used by TCP sockets.
628 It is guaranteed to each TCP socket, even under moderate memory
633 default: initial size of receive buffer used by TCP sockets.
634 This value overrides net.core.rmem_default used by other protocols.
635 Default: 131072 bytes.
636 This value results in initial window of 65535.
638 max: maximal size of receive buffer allowed for automatically
639 selected receiver buffers for TCP socket. This value does not override
640 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
641 automatic tuning of that socket's receive buffer size, in which
642 case this value is ignored.
643 Default: between 131072 and 6MB, depending on RAM size.
646 Enable select acknowledgments (SACKS).
648 tcp_comp_sack_delay_ns - LONG INTEGER
649 TCP tries to reduce number of SACK sent, using a timer
650 based on 5% of SRTT, capped by this sysctl, in nano seconds.
651 The default is 1ms, based on TSO autosizing period.
653 Default : 1,000,000 ns (1 ms)
655 tcp_comp_sack_slack_ns - LONG INTEGER
656 This sysctl control the slack used when arming the
657 timer used by SACK compression. This gives extra time
658 for small RTT flows, and reduces system overhead by allowing
659 opportunistic reduction of timer interrupts.
661 Default : 100,000 ns (100 us)
663 tcp_comp_sack_nr - INTEGER
664 Max number of SACK that can be compressed.
665 Using 0 disables SACK compression.
669 tcp_slow_start_after_idle - BOOLEAN
670 If set, provide RFC2861 behavior and time out the congestion
671 window after an idle period. An idle period is defined at
672 the current RTO. If unset, the congestion window will not
673 be timed out after an idle period.
678 Use the Host requirements interpretation of the TCP urgent pointer field.
679 Most hosts use the older BSD interpretation, so if you turn this on
680 Linux might not communicate correctly with them.
684 tcp_synack_retries - INTEGER
685 Number of times SYNACKs for a passive TCP connection attempt will
686 be retransmitted. Should not be higher than 255. Default value
687 is 5, which corresponds to 31seconds till the last retransmission
688 with the current initial RTO of 1second. With this the final timeout
689 for a passive TCP connection will happen after 63seconds.
691 tcp_syncookies - INTEGER
692 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
693 Send out syncookies when the syn backlog queue of a socket
694 overflows. This is to prevent against the common 'SYN flood attack'
697 Note, that syncookies is fallback facility.
698 It MUST NOT be used to help highly loaded servers to stand
699 against legal connection rate. If you see SYN flood warnings
700 in your logs, but investigation shows that they occur
701 because of overload with legal connections, you should tune
702 another parameters until this warning disappear.
703 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
705 syncookies seriously violate TCP protocol, do not allow
706 to use TCP extensions, can result in serious degradation
707 of some services (f.e. SMTP relaying), visible not by you,
708 but your clients and relays, contacting you. While you see
709 SYN flood warnings in logs not being really flooded, your server
710 is seriously misconfigured.
712 If you want to test which effects syncookies have to your
713 network connections you can set this knob to 2 to enable
714 unconditionally generation of syncookies.
716 tcp_migrate_req - BOOLEAN
717 The incoming connection is tied to a specific listening socket when
718 the initial SYN packet is received during the three-way handshake.
719 When a listener is closed, in-flight request sockets during the
720 handshake and established sockets in the accept queue are aborted.
722 If the listener has SO_REUSEPORT enabled, other listeners on the
723 same port should have been able to accept such connections. This
724 option makes it possible to migrate such child sockets to another
725 listener after close() or shutdown().
727 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
728 usually be used to define the policy to pick an alive listener.
729 Otherwise, the kernel will randomly pick an alive listener only if
730 this option is enabled.
732 Note that migration between listeners with different settings may
733 crash applications. Let's say migration happens from listener A to
734 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
735 the requests migrated from A. To avoid such a situation, cancel
736 migration by returning SK_DROP in the type of eBPF program, or
741 tcp_fastopen - INTEGER
742 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
745 The client support is enabled by flag 0x1 (on by default). The client
746 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
747 rather than connect() to send data in SYN.
749 The server support is enabled by flag 0x2 (off by default). Then
750 either enable for all listeners with another flag (0x400) or
751 enable individual listeners via TCP_FASTOPEN socket option with
752 the option value being the length of the syn-data backlog.
754 The values (bitmap) are
756 ===== ======== ======================================================
757 0x1 (client) enables sending data in the opening SYN on the client.
758 0x2 (server) enables the server support, i.e., allowing data in
759 a SYN packet to be accepted and passed to the
760 application before 3-way handshake finishes.
761 0x4 (client) send data in the opening SYN regardless of cookie
762 availability and without a cookie option.
763 0x200 (server) accept data-in-SYN w/o any cookie option present.
764 0x400 (server) enable all listeners to support Fast Open by
765 default without explicit TCP_FASTOPEN socket option.
766 ===== ======== ======================================================
770 Note that additional client or server features are only
771 effective if the basic support (0x1 and 0x2) are enabled respectively.
773 tcp_fastopen_blackhole_timeout_sec - INTEGER
774 Initial time period in second to disable Fastopen on active TCP sockets
775 when a TFO firewall blackhole issue happens.
776 This time period will grow exponentially when more blackhole issues
777 get detected right after Fastopen is re-enabled and will reset to
778 initial value when the blackhole issue goes away.
779 0 to disable the blackhole detection.
781 By default, it is set to 0 (feature is disabled).
783 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
784 The list consists of a primary key and an optional backup key. The
785 primary key is used for both creating and validating cookies, while the
786 optional backup key is only used for validating cookies. The purpose of
787 the backup key is to maximize TFO validation when keys are rotated.
789 A randomly chosen primary key may be configured by the kernel if
790 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
791 TCP_FASTOPEN setsockopt() optname is set and a key has not been
792 previously configured via sysctl. If keys are configured via
793 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
794 per-socket keys will be used instead of any keys that are specified via
797 A key is specified as 4 8-digit hexadecimal integers which are separated
798 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
799 omitted. A primary and a backup key may be specified by separating them
800 by a comma. If only one key is specified, it becomes the primary key and
801 any previously configured backup keys are removed.
803 tcp_syn_retries - INTEGER
804 Number of times initial SYNs for an active TCP connection attempt
805 will be retransmitted. Should not be higher than 127. Default value
806 is 6, which corresponds to 63seconds till the last retransmission
807 with the current initial RTO of 1second. With this the final timeout
808 for an active TCP connection attempt will happen after 127seconds.
810 tcp_timestamps - INTEGER
811 Enable timestamps as defined in RFC1323.
814 - 1: Enable timestamps as defined in RFC1323 and use random offset for
815 each connection rather than only using the current time.
816 - 2: Like 1, but without random offsets.
820 tcp_min_tso_segs - INTEGER
821 Minimal number of segments per TSO frame.
823 Since linux-3.12, TCP does an automatic sizing of TSO frames,
824 depending on flow rate, instead of filling 64Kbytes packets.
825 For specific usages, it's possible to force TCP to build big
826 TSO frames. Note that TCP stack might split too big TSO packets
827 if available window is too small.
831 tcp_pacing_ss_ratio - INTEGER
832 sk->sk_pacing_rate is set by TCP stack using a ratio applied
833 to current rate. (current_rate = cwnd * mss / srtt)
834 If TCP is in slow start, tcp_pacing_ss_ratio is applied
835 to let TCP probe for bigger speeds, assuming cwnd can be
836 doubled every other RTT.
840 tcp_pacing_ca_ratio - INTEGER
841 sk->sk_pacing_rate is set by TCP stack using a ratio applied
842 to current rate. (current_rate = cwnd * mss / srtt)
843 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
844 is applied to conservatively probe for bigger throughput.
848 tcp_tso_win_divisor - INTEGER
849 This allows control over what percentage of the congestion window
850 can be consumed by a single TSO frame.
851 The setting of this parameter is a choice between burstiness and
852 building larger TSO frames.
856 tcp_tw_reuse - INTEGER
857 Enable reuse of TIME-WAIT sockets for new connections when it is
858 safe from protocol viewpoint.
862 - 2 - enable for loopback traffic only
864 It should not be changed without advice/request of technical
869 tcp_window_scaling - BOOLEAN
870 Enable window scaling as defined in RFC1323.
872 tcp_wmem - vector of 3 INTEGERs: min, default, max
873 min: Amount of memory reserved for send buffers for TCP sockets.
874 Each TCP socket has rights to use it due to fact of its birth.
878 default: initial size of send buffer used by TCP sockets. This
879 value overrides net.core.wmem_default used by other protocols.
881 It is usually lower than net.core.wmem_default.
885 max: Maximal amount of memory allowed for automatically tuned
886 send buffers for TCP sockets. This value does not override
887 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
888 automatic tuning of that socket's send buffer size, in which case
889 this value is ignored.
891 Default: between 64K and 4MB, depending on RAM size.
893 tcp_notsent_lowat - UNSIGNED INTEGER
894 A TCP socket can control the amount of unsent bytes in its write queue,
895 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
896 reports POLLOUT events if the amount of unsent bytes is below a per
897 socket value, and if the write queue is not full. sendmsg() will
898 also not add new buffers if the limit is hit.
900 This global variable controls the amount of unsent data for
901 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
902 to the global variable has immediate effect.
904 Default: UINT_MAX (0xFFFFFFFF)
906 tcp_workaround_signed_windows - BOOLEAN
907 If set, assume no receipt of a window scaling option means the
908 remote TCP is broken and treats the window as a signed quantity.
909 If unset, assume the remote TCP is not broken even if we do
910 not receive a window scaling option from them.
914 tcp_thin_linear_timeouts - BOOLEAN
915 Enable dynamic triggering of linear timeouts for thin streams.
916 If set, a check is performed upon retransmission by timeout to
917 determine if the stream is thin (less than 4 packets in flight).
918 As long as the stream is found to be thin, up to 6 linear
919 timeouts may be performed before exponential backoff mode is
920 initiated. This improves retransmission latency for
921 non-aggressive thin streams, often found to be time-dependent.
922 For more information on thin streams, see
923 Documentation/networking/tcp-thin.rst
927 tcp_limit_output_bytes - INTEGER
928 Controls TCP Small Queue limit per tcp socket.
929 TCP bulk sender tends to increase packets in flight until it
930 gets losses notifications. With SNDBUF autotuning, this can
931 result in a large amount of packets queued on the local machine
932 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
933 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
934 limits the number of bytes on qdisc or device to reduce artificial
935 RTT/cwnd and reduce bufferbloat.
937 Default: 1048576 (16 * 65536)
939 tcp_challenge_ack_limit - INTEGER
940 Limits number of Challenge ACK sent per second, as recommended
941 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
944 tcp_rx_skb_cache - BOOLEAN
945 Controls a per TCP socket cache of one skb, that might help
946 performance of some workloads. This might be dangerous
947 on systems with a lot of TCP sockets, since it increases
950 Default: 0 (disabled)
955 udp_l3mdev_accept - BOOLEAN
956 Enabling this option allows a "global" bound socket to work
957 across L3 master domains (e.g., VRFs) with packets capable of
958 being received regardless of the L3 domain in which they
959 originated. Only valid when the kernel was compiled with
960 CONFIG_NET_L3_MASTER_DEV.
962 Default: 0 (disabled)
964 udp_mem - vector of 3 INTEGERs: min, pressure, max
965 Number of pages allowed for queueing by all UDP sockets.
967 min: Below this number of pages UDP is not bothered about its
968 memory appetite. When amount of memory allocated by UDP exceeds
969 this number, UDP starts to moderate memory usage.
971 pressure: This value was introduced to follow format of tcp_mem.
973 max: Number of pages allowed for queueing by all UDP sockets.
975 Default is calculated at boot time from amount of available memory.
977 udp_rmem_min - INTEGER
978 Minimal size of receive buffer used by UDP sockets in moderation.
979 Each UDP socket is able to use the size for receiving data, even if
980 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
984 udp_wmem_min - INTEGER
985 Minimal size of send buffer used by UDP sockets in moderation.
986 Each UDP socket is able to use the size for sending data, even if
987 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
994 raw_l3mdev_accept - BOOLEAN
995 Enabling this option allows a "global" bound socket to work
996 across L3 master domains (e.g., VRFs) with packets capable of
997 being received regardless of the L3 domain in which they
998 originated. Only valid when the kernel was compiled with
999 CONFIG_NET_L3_MASTER_DEV.
1001 Default: 1 (enabled)
1006 cipso_cache_enable - BOOLEAN
1007 If set, enable additions to and lookups from the CIPSO label mapping
1008 cache. If unset, additions are ignored and lookups always result in a
1009 miss. However, regardless of the setting the cache is still
1010 invalidated when required when means you can safely toggle this on and
1011 off and the cache will always be "safe".
1015 cipso_cache_bucket_size - INTEGER
1016 The CIPSO label cache consists of a fixed size hash table with each
1017 hash bucket containing a number of cache entries. This variable limits
1018 the number of entries in each hash bucket; the larger the value is, the
1019 more CIPSO label mappings that can be cached. When the number of
1020 entries in a given hash bucket reaches this limit adding new entries
1021 causes the oldest entry in the bucket to be removed to make room.
1025 cipso_rbm_optfmt - BOOLEAN
1026 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1027 the CIPSO draft specification (see Documentation/netlabel for details).
1028 This means that when set the CIPSO tag will be padded with empty
1029 categories in order to make the packet data 32-bit aligned.
1033 cipso_rbm_structvalid - BOOLEAN
1034 If set, do a very strict check of the CIPSO option when
1035 ip_options_compile() is called. If unset, relax the checks done during
1036 ip_options_compile(). Either way is "safe" as errors are caught else
1037 where in the CIPSO processing code but setting this to 0 (False) should
1038 result in less work (i.e. it should be faster) but could cause problems
1039 with other implementations that require strict checking.
1046 ip_local_port_range - 2 INTEGERS
1047 Defines the local port range that is used by TCP and UDP to
1048 choose the local port. The first number is the first, the
1049 second the last local port number.
1050 If possible, it is better these numbers have different parity
1051 (one even and one odd value).
1052 Must be greater than or equal to ip_unprivileged_port_start.
1053 The default values are 32768 and 60999 respectively.
1055 ip_local_reserved_ports - list of comma separated ranges
1056 Specify the ports which are reserved for known third-party
1057 applications. These ports will not be used by automatic port
1058 assignments (e.g. when calling connect() or bind() with port
1059 number 0). Explicit port allocation behavior is unchanged.
1061 The format used for both input and output is a comma separated
1062 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1063 10). Writing to the file will clear all previously reserved
1064 ports and update the current list with the one given in the
1067 Note that ip_local_port_range and ip_local_reserved_ports
1068 settings are independent and both are considered by the kernel
1069 when determining which ports are available for automatic port
1072 You can reserve ports which are not in the current
1073 ip_local_port_range, e.g.::
1075 $ cat /proc/sys/net/ipv4/ip_local_port_range
1077 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1080 although this is redundant. However such a setting is useful
1081 if later the port range is changed to a value that will
1082 include the reserved ports.
1086 ip_unprivileged_port_start - INTEGER
1087 This is a per-namespace sysctl. It defines the first
1088 unprivileged port in the network namespace. Privileged ports
1089 require root or CAP_NET_BIND_SERVICE in order to bind to them.
1090 To disable all privileged ports, set this to 0. They must not
1091 overlap with the ip_local_port_range.
1095 ip_nonlocal_bind - BOOLEAN
1096 If set, allows processes to bind() to non-local IP addresses,
1097 which can be quite useful - but may break some applications.
1101 ip_autobind_reuse - BOOLEAN
1102 By default, bind() does not select the ports automatically even if
1103 the new socket and all sockets bound to the port have SO_REUSEADDR.
1104 ip_autobind_reuse allows bind() to reuse the port and this is useful
1105 when you use bind()+connect(), but may break some applications.
1106 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1107 option should only be set by experts.
1110 ip_dynaddr - INTEGER
1111 If set non-zero, enables support for dynamic addresses.
1112 If set to a non-zero value larger than 1, a kernel log
1113 message will be printed when dynamic address rewriting
1118 ip_early_demux - BOOLEAN
1119 Optimize input packet processing down to one demux for
1120 certain kinds of local sockets. Currently we only do this
1121 for established TCP and connected UDP sockets.
1123 It may add an additional cost for pure routing workloads that
1124 reduces overall throughput, in such case you should disable it.
1128 ping_group_range - 2 INTEGERS
1129 Restrict ICMP_PROTO datagram sockets to users in the group range.
1130 The default is "1 0", meaning, that nobody (not even root) may
1131 create ping sockets. Setting it to "100 100" would grant permissions
1132 to the single group. "0 4294967295" would enable it for the world, "100
1133 4294967295" would enable it for the users, but not daemons.
1135 tcp_early_demux - BOOLEAN
1136 Enable early demux for established TCP sockets.
1140 udp_early_demux - BOOLEAN
1141 Enable early demux for connected UDP sockets. Disable this if
1142 your system could experience more unconnected load.
1146 icmp_echo_ignore_all - BOOLEAN
1147 If set non-zero, then the kernel will ignore all ICMP ECHO
1148 requests sent to it.
1152 icmp_echo_ignore_broadcasts - BOOLEAN
1153 If set non-zero, then the kernel will ignore all ICMP ECHO and
1154 TIMESTAMP requests sent to it via broadcast/multicast.
1158 icmp_ratelimit - INTEGER
1159 Limit the maximal rates for sending ICMP packets whose type matches
1160 icmp_ratemask (see below) to specific targets.
1161 0 to disable any limiting,
1162 otherwise the minimal space between responses in milliseconds.
1163 Note that another sysctl, icmp_msgs_per_sec limits the number
1164 of ICMP packets sent on all targets.
1168 icmp_msgs_per_sec - INTEGER
1169 Limit maximal number of ICMP packets sent per second from this host.
1170 Only messages whose type matches icmp_ratemask (see below) are
1171 controlled by this limit. For security reasons, the precise count
1172 of messages per second is randomized.
1176 icmp_msgs_burst - INTEGER
1177 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1178 while icmp_msgs_burst controls the burst size of these packets.
1179 For security reasons, the precise burst size is randomized.
1183 icmp_ratemask - INTEGER
1184 Mask made of ICMP types for which rates are being limited.
1186 Significant bits: IHGFEDCBA9876543210
1188 Default mask: 0000001100000011000 (6168)
1190 Bit definitions (see include/linux/icmp.h):
1192 = =========================
1194 3 Destination Unreachable [1]_
1195 4 Source Quench [1]_
1198 B Time Exceeded [1]_
1199 C Parameter Problem [1]_
1204 H Address Mask Request
1205 I Address Mask Reply
1206 = =========================
1208 .. [1] These are rate limited by default (see default mask above)
1210 icmp_ignore_bogus_error_responses - BOOLEAN
1211 Some routers violate RFC1122 by sending bogus responses to broadcast
1212 frames. Such violations are normally logged via a kernel warning.
1213 If this is set to TRUE, the kernel will not give such warnings, which
1214 will avoid log file clutter.
1218 icmp_errors_use_inbound_ifaddr - BOOLEAN
1220 If zero, icmp error messages are sent with the primary address of
1221 the exiting interface.
1223 If non-zero, the message will be sent with the primary address of
1224 the interface that received the packet that caused the icmp error.
1225 This is the behaviour network many administrators will expect from
1226 a router. And it can make debugging complicated network layouts
1229 Note that if no primary address exists for the interface selected,
1230 then the primary address of the first non-loopback interface that
1231 has one will be used regardless of this setting.
1235 igmp_max_memberships - INTEGER
1236 Change the maximum number of multicast groups we can subscribe to.
1239 Theoretical maximum value is bounded by having to send a membership
1240 report in a single datagram (i.e. the report can't span multiple
1241 datagrams, or risk confusing the switch and leaving groups you don't
1244 The number of supported groups 'M' is bounded by the number of group
1245 report entries you can fit into a single datagram of 65535 bytes.
1247 M = 65536-sizeof (ip header)/(sizeof(Group record))
1249 Group records are variable length, with a minimum of 12 bytes.
1250 So net.ipv4.igmp_max_memberships should not be set higher than:
1252 (65536-24) / 12 = 5459
1254 The value 5459 assumes no IP header options, so in practice
1255 this number may be lower.
1257 igmp_max_msf - INTEGER
1258 Maximum number of addresses allowed in the source filter list for a
1264 Controls the IGMP query robustness variable (see RFC2236 8.1).
1266 Default: 2 (as specified by RFC2236 8.1)
1268 Minimum: 1 (as specified by RFC6636 4.5)
1270 force_igmp_version - INTEGER
1271 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1272 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1273 Present timer expires.
1274 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1275 receive IGMPv2/v3 query.
1276 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1277 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1278 - 3 - Enforce to use IGMP version 3. The same react with default 0.
1282 this is not the same with force_mld_version because IGMPv3 RFC3376
1283 Security Considerations does not have clear description that we could
1284 ignore other version messages completely as MLDv2 RFC3810. So make
1285 this value as default 0 is recommended.
1287 ``conf/interface/*``
1288 changes special settings per interface (where
1289 interface" is the name of your network interface)
1292 is special, changes the settings for all interfaces
1294 log_martians - BOOLEAN
1295 Log packets with impossible addresses to kernel log.
1296 log_martians for the interface will be enabled if at least one of
1297 conf/{all,interface}/log_martians is set to TRUE,
1298 it will be disabled otherwise
1300 accept_redirects - BOOLEAN
1301 Accept ICMP redirect messages.
1302 accept_redirects for the interface will be enabled if:
1304 - both conf/{all,interface}/accept_redirects are TRUE in the case
1305 forwarding for the interface is enabled
1309 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1310 case forwarding for the interface is disabled
1312 accept_redirects for the interface will be disabled otherwise
1319 forwarding - BOOLEAN
1320 Enable IP forwarding on this interface. This controls whether packets
1321 received _on_ this interface can be forwarded.
1323 mc_forwarding - BOOLEAN
1324 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1325 and a multicast routing daemon is required.
1326 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1327 routing for the interface
1330 Integer value used to differentiate the devices by the medium they
1331 are attached to. Two devices can have different id values when
1332 the broadcast packets are received only on one of them.
1333 The default value 0 means that the device is the only interface
1334 to its medium, value of -1 means that medium is not known.
1336 Currently, it is used to change the proxy_arp behavior:
1337 the proxy_arp feature is enabled for packets forwarded between
1338 two devices attached to different media.
1343 proxy_arp for the interface will be enabled if at least one of
1344 conf/{all,interface}/proxy_arp is set to TRUE,
1345 it will be disabled otherwise
1347 proxy_arp_pvlan - BOOLEAN
1348 Private VLAN proxy arp.
1350 Basically allow proxy arp replies back to the same interface
1351 (from which the ARP request/solicitation was received).
1353 This is done to support (ethernet) switch features, like RFC
1354 3069, where the individual ports are NOT allowed to
1355 communicate with each other, but they are allowed to talk to
1356 the upstream router. As described in RFC 3069, it is possible
1357 to allow these hosts to communicate through the upstream
1358 router by proxy_arp'ing. Don't need to be used together with
1361 This technology is known by different names:
1363 In RFC 3069 it is called VLAN Aggregation.
1364 Cisco and Allied Telesyn call it Private VLAN.
1365 Hewlett-Packard call it Source-Port filtering or port-isolation.
1366 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1368 shared_media - BOOLEAN
1369 Send(router) or accept(host) RFC1620 shared media redirects.
1370 Overrides secure_redirects.
1372 shared_media for the interface will be enabled if at least one of
1373 conf/{all,interface}/shared_media is set to TRUE,
1374 it will be disabled otherwise
1378 secure_redirects - BOOLEAN
1379 Accept ICMP redirect messages only to gateways listed in the
1380 interface's current gateway list. Even if disabled, RFC1122 redirect
1383 Overridden by shared_media.
1385 secure_redirects for the interface will be enabled if at least one of
1386 conf/{all,interface}/secure_redirects is set to TRUE,
1387 it will be disabled otherwise
1391 send_redirects - BOOLEAN
1392 Send redirects, if router.
1394 send_redirects for the interface will be enabled if at least one of
1395 conf/{all,interface}/send_redirects is set to TRUE,
1396 it will be disabled otherwise
1400 bootp_relay - BOOLEAN
1401 Accept packets with source address 0.b.c.d destined
1402 not to this host as local ones. It is supposed, that
1403 BOOTP relay daemon will catch and forward such packets.
1404 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1409 Not Implemented Yet.
1411 accept_source_route - BOOLEAN
1412 Accept packets with SRR option.
1413 conf/all/accept_source_route must also be set to TRUE to accept packets
1414 with SRR option on the interface
1421 accept_local - BOOLEAN
1422 Accept packets with local source addresses. In combination with
1423 suitable routing, this can be used to direct packets between two
1424 local interfaces over the wire and have them accepted properly.
1427 route_localnet - BOOLEAN
1428 Do not consider loopback addresses as martian source or destination
1429 while routing. This enables the use of 127/8 for local routing purposes.
1434 - 0 - No source validation.
1435 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1436 Each incoming packet is tested against the FIB and if the interface
1437 is not the best reverse path the packet check will fail.
1438 By default failed packets are discarded.
1439 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1440 Each incoming packet's source address is also tested against the FIB
1441 and if the source address is not reachable via any interface
1442 the packet check will fail.
1444 Current recommended practice in RFC3704 is to enable strict mode
1445 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1446 or other complicated routing, then loose mode is recommended.
1448 The max value from conf/{all,interface}/rp_filter is used
1449 when doing source validation on the {interface}.
1451 Default value is 0. Note that some distributions enable it
1454 arp_filter - BOOLEAN
1455 - 1 - Allows you to have multiple network interfaces on the same
1456 subnet, and have the ARPs for each interface be answered
1457 based on whether or not the kernel would route a packet from
1458 the ARP'd IP out that interface (therefore you must use source
1459 based routing for this to work). In other words it allows control
1460 of which cards (usually 1) will respond to an arp request.
1462 - 0 - (default) The kernel can respond to arp requests with addresses
1463 from other interfaces. This may seem wrong but it usually makes
1464 sense, because it increases the chance of successful communication.
1465 IP addresses are owned by the complete host on Linux, not by
1466 particular interfaces. Only for more complex setups like load-
1467 balancing, does this behaviour cause problems.
1469 arp_filter for the interface will be enabled if at least one of
1470 conf/{all,interface}/arp_filter is set to TRUE,
1471 it will be disabled otherwise
1473 arp_announce - INTEGER
1474 Define different restriction levels for announcing the local
1475 source IP address from IP packets in ARP requests sent on
1478 - 0 - (default) Use any local address, configured on any interface
1479 - 1 - Try to avoid local addresses that are not in the target's
1480 subnet for this interface. This mode is useful when target
1481 hosts reachable via this interface require the source IP
1482 address in ARP requests to be part of their logical network
1483 configured on the receiving interface. When we generate the
1484 request we will check all our subnets that include the
1485 target IP and will preserve the source address if it is from
1486 such subnet. If there is no such subnet we select source
1487 address according to the rules for level 2.
1488 - 2 - Always use the best local address for this target.
1489 In this mode we ignore the source address in the IP packet
1490 and try to select local address that we prefer for talks with
1491 the target host. Such local address is selected by looking
1492 for primary IP addresses on all our subnets on the outgoing
1493 interface that include the target IP address. If no suitable
1494 local address is found we select the first local address
1495 we have on the outgoing interface or on all other interfaces,
1496 with the hope we will receive reply for our request and
1497 even sometimes no matter the source IP address we announce.
1499 The max value from conf/{all,interface}/arp_announce is used.
1501 Increasing the restriction level gives more chance for
1502 receiving answer from the resolved target while decreasing
1503 the level announces more valid sender's information.
1505 arp_ignore - INTEGER
1506 Define different modes for sending replies in response to
1507 received ARP requests that resolve local target IP addresses:
1509 - 0 - (default): reply for any local target IP address, configured
1511 - 1 - reply only if the target IP address is local address
1512 configured on the incoming interface
1513 - 2 - reply only if the target IP address is local address
1514 configured on the incoming interface and both with the
1515 sender's IP address are part from same subnet on this interface
1516 - 3 - do not reply for local addresses configured with scope host,
1517 only resolutions for global and link addresses are replied
1519 - 8 - do not reply for all local addresses
1521 The max value from conf/{all,interface}/arp_ignore is used
1522 when ARP request is received on the {interface}
1524 arp_notify - BOOLEAN
1525 Define mode for notification of address and device changes.
1527 == ==========================================================
1528 0 (default): do nothing
1529 1 Generate gratuitous arp requests when device is brought up
1530 or hardware address changes.
1531 == ==========================================================
1533 arp_accept - BOOLEAN
1534 Define behavior for gratuitous ARP frames who's IP is not
1535 already present in the ARP table:
1537 - 0 - don't create new entries in the ARP table
1538 - 1 - create new entries in the ARP table
1540 Both replies and requests type gratuitous arp will trigger the
1541 ARP table to be updated, if this setting is on.
1543 If the ARP table already contains the IP address of the
1544 gratuitous arp frame, the arp table will be updated regardless
1545 if this setting is on or off.
1547 mcast_solicit - INTEGER
1548 The maximum number of multicast probes in INCOMPLETE state,
1549 when the associated hardware address is unknown. Defaults
1552 ucast_solicit - INTEGER
1553 The maximum number of unicast probes in PROBE state, when
1554 the hardware address is being reconfirmed. Defaults to 3.
1556 app_solicit - INTEGER
1557 The maximum number of probes to send to the user space ARP daemon
1558 via netlink before dropping back to multicast probes (see
1559 mcast_resolicit). Defaults to 0.
1561 mcast_resolicit - INTEGER
1562 The maximum number of multicast probes after unicast and
1563 app probes in PROBE state. Defaults to 0.
1565 disable_policy - BOOLEAN
1566 Disable IPSEC policy (SPD) for this interface
1568 disable_xfrm - BOOLEAN
1569 Disable IPSEC encryption on this interface, whatever the policy
1571 igmpv2_unsolicited_report_interval - INTEGER
1572 The interval in milliseconds in which the next unsolicited
1573 IGMPv1 or IGMPv2 report retransmit will take place.
1575 Default: 10000 (10 seconds)
1577 igmpv3_unsolicited_report_interval - INTEGER
1578 The interval in milliseconds in which the next unsolicited
1579 IGMPv3 report retransmit will take place.
1581 Default: 1000 (1 seconds)
1583 promote_secondaries - BOOLEAN
1584 When a primary IP address is removed from this interface
1585 promote a corresponding secondary IP address instead of
1586 removing all the corresponding secondary IP addresses.
1588 drop_unicast_in_l2_multicast - BOOLEAN
1589 Drop any unicast IP packets that are received in link-layer
1590 multicast (or broadcast) frames.
1592 This behavior (for multicast) is actually a SHOULD in RFC
1593 1122, but is disabled by default for compatibility reasons.
1597 drop_gratuitous_arp - BOOLEAN
1598 Drop all gratuitous ARP frames, for example if there's a known
1599 good ARP proxy on the network and such frames need not be used
1600 (or in the case of 802.11, must not be used to prevent attacks.)
1606 Allows you to write a number, which can be used as required.
1610 xfrm4_gc_thresh - INTEGER
1611 (Obsolete since linux-4.14)
1612 The threshold at which we will start garbage collecting for IPv4
1613 destination cache entries. At twice this value the system will
1614 refuse new allocations.
1616 igmp_link_local_mcast_reports - BOOLEAN
1617 Enable IGMP reports for link local multicast groups in the
1623 kuznet@ms2.inr.ac.ru
1630 delon.nicolas@wanadoo.fr
1635 /proc/sys/net/ipv6/* Variables
1636 ==============================
1638 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1639 apply to IPv6 [XXX?].
1641 bindv6only - BOOLEAN
1642 Default value for IPV6_V6ONLY socket option,
1643 which restricts use of the IPv6 socket to IPv6 communication
1646 - TRUE: disable IPv4-mapped address feature
1647 - FALSE: enable IPv4-mapped address feature
1649 Default: FALSE (as specified in RFC3493)
1651 flowlabel_consistency - BOOLEAN
1652 Protect the consistency (and unicity) of flow label.
1653 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1661 auto_flowlabels - INTEGER
1662 Automatically generate flow labels based on a flow hash of the
1663 packet. This allows intermediate devices, such as routers, to
1664 identify packet flows for mechanisms like Equal Cost Multipath
1665 Routing (see RFC 6438).
1667 = ===========================================================
1668 0 automatic flow labels are completely disabled
1669 1 automatic flow labels are enabled by default, they can be
1670 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1672 2 automatic flow labels are allowed, they may be enabled on a
1673 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1674 3 automatic flow labels are enabled and enforced, they cannot
1675 be disabled by the socket option
1676 = ===========================================================
1680 flowlabel_state_ranges - BOOLEAN
1681 Split the flow label number space into two ranges. 0-0x7FFFF is
1682 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1683 is reserved for stateless flow labels as described in RFC6437.
1690 flowlabel_reflect - INTEGER
1691 Control flow label reflection. Needed for Path MTU
1692 Discovery to work with Equal Cost Multipath Routing in anycast
1693 environments. See RFC 7690 and:
1694 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1698 - 1: enabled for established flows
1700 Note that this prevents automatic flowlabel changes, as done
1701 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
1702 and "tcp: Change txhash on every SYN and RTO retransmit"
1704 - 2: enabled for TCP RESET packets (no active listener)
1705 If set, a RST packet sent in response to a SYN packet on a closed
1706 port will reflect the incoming flow label.
1708 - 4: enabled for ICMPv6 echo reply messages.
1712 fib_multipath_hash_policy - INTEGER
1713 Controls which hash policy to use for multipath routes.
1715 Default: 0 (Layer 3)
1719 - 0 - Layer 3 (source and destination addresses plus flow label)
1720 - 1 - Layer 4 (standard 5-tuple)
1721 - 2 - Layer 3 or inner Layer 3 if present
1723 anycast_src_echo_reply - BOOLEAN
1724 Controls the use of anycast addresses as source addresses for ICMPv6
1732 idgen_delay - INTEGER
1733 Controls the delay in seconds after which time to retry
1734 privacy stable address generation if a DAD conflict is
1737 Default: 1 (as specified in RFC7217)
1739 idgen_retries - INTEGER
1740 Controls the number of retries to generate a stable privacy
1741 address if a DAD conflict is detected.
1743 Default: 3 (as specified in RFC7217)
1746 Controls the MLD query robustness variable (see RFC3810 9.1).
1748 Default: 2 (as specified by RFC3810 9.1)
1750 Minimum: 1 (as specified by RFC6636 4.5)
1752 max_dst_opts_number - INTEGER
1753 Maximum number of non-padding TLVs allowed in a Destination
1754 options extension header. If this value is less than zero
1755 then unknown options are disallowed and the number of known
1756 TLVs allowed is the absolute value of this number.
1760 max_hbh_opts_number - INTEGER
1761 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1762 options extension header. If this value is less than zero
1763 then unknown options are disallowed and the number of known
1764 TLVs allowed is the absolute value of this number.
1768 max_dst_opts_length - INTEGER
1769 Maximum length allowed for a Destination options extension
1772 Default: INT_MAX (unlimited)
1774 max_hbh_length - INTEGER
1775 Maximum length allowed for a Hop-by-Hop options extension
1778 Default: INT_MAX (unlimited)
1780 skip_notify_on_dev_down - BOOLEAN
1781 Controls whether an RTM_DELROUTE message is generated for routes
1782 removed when a device is taken down or deleted. IPv4 does not
1783 generate this message; IPv6 does by default. Setting this sysctl
1784 to true skips the message, making IPv4 and IPv6 on par in relying
1785 on userspace caches to track link events and evict routes.
1787 Default: false (generate message)
1789 nexthop_compat_mode - BOOLEAN
1790 New nexthop API provides a means for managing nexthops independent of
1791 prefixes. Backwards compatibilty with old route format is enabled by
1792 default which means route dumps and notifications contain the new
1793 nexthop attribute but also the full, expanded nexthop definition.
1794 Further, updates or deletes of a nexthop configuration generate route
1795 notifications for each fib entry using the nexthop. Once a system
1796 understands the new API, this sysctl can be disabled to achieve full
1797 performance benefits of the new API by disabling the nexthop expansion
1798 and extraneous notifications.
1799 Default: true (backward compat mode)
1803 ip6frag_high_thresh - INTEGER
1804 Maximum memory used to reassemble IPv6 fragments. When
1805 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1806 the fragment handler will toss packets until ip6frag_low_thresh
1809 ip6frag_low_thresh - INTEGER
1810 See ip6frag_high_thresh
1812 ip6frag_time - INTEGER
1813 Time in seconds to keep an IPv6 fragment in memory.
1815 IPv6 Segment Routing:
1817 seg6_flowlabel - INTEGER
1818 Controls the behaviour of computing the flowlabel of outer
1819 IPv6 header in case of SR T.encaps
1821 == =======================================================
1822 -1 set flowlabel to zero.
1823 0 copy flowlabel from Inner packet in case of Inner IPv6
1824 (Set flowlabel to 0 in case IPv4/L2)
1825 1 Compute the flowlabel using seg6_make_flowlabel()
1826 == =======================================================
1831 Change the interface-specific default settings.
1835 Change all the interface-specific settings.
1837 [XXX: Other special features than forwarding?]
1839 conf/all/forwarding - BOOLEAN
1840 Enable global IPv6 forwarding between all interfaces.
1842 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1843 to control which interfaces may forward packets and which not.
1845 This also sets all interfaces' Host/Router setting
1846 'forwarding' to the specified value. See below for details.
1848 This referred to as global forwarding.
1853 fwmark_reflect - BOOLEAN
1854 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1855 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1856 If unset, these packets have a fwmark of zero. If set, they have the
1857 fwmark of the packet they are replying to.
1861 ``conf/interface/*``:
1862 Change special settings per interface.
1864 The functional behaviour for certain settings is different
1865 depending on whether local forwarding is enabled or not.
1868 Accept Router Advertisements; autoconfigure using them.
1870 It also determines whether or not to transmit Router
1871 Solicitations. If and only if the functional setting is to
1872 accept Router Advertisements, Router Solicitations will be
1875 Possible values are:
1877 == ===========================================================
1878 0 Do not accept Router Advertisements.
1879 1 Accept Router Advertisements if forwarding is disabled.
1880 2 Overrule forwarding behaviour. Accept Router Advertisements
1881 even if forwarding is enabled.
1882 == ===========================================================
1886 - enabled if local forwarding is disabled.
1887 - disabled if local forwarding is enabled.
1889 accept_ra_defrtr - BOOLEAN
1890 Learn default router in Router Advertisement.
1894 - enabled if accept_ra is enabled.
1895 - disabled if accept_ra is disabled.
1897 accept_ra_from_local - BOOLEAN
1898 Accept RA with source-address that is found on local machine
1899 if the RA is otherwise proper and able to be accepted.
1901 Default is to NOT accept these as it may be an un-intended
1906 - enabled if accept_ra_from_local is enabled
1907 on a specific interface.
1908 - disabled if accept_ra_from_local is disabled
1909 on a specific interface.
1911 accept_ra_min_hop_limit - INTEGER
1912 Minimum hop limit Information in Router Advertisement.
1914 Hop limit Information in Router Advertisement less than this
1915 variable shall be ignored.
1919 accept_ra_min_lft - INTEGER
1920 Minimum acceptable lifetime value in Router Advertisement.
1922 RA sections with a lifetime less than this value shall be
1923 ignored. Zero lifetimes stay unaffected.
1927 accept_ra_pinfo - BOOLEAN
1928 Learn Prefix Information in Router Advertisement.
1932 - enabled if accept_ra is enabled.
1933 - disabled if accept_ra is disabled.
1935 accept_ra_rt_info_min_plen - INTEGER
1936 Minimum prefix length of Route Information in RA.
1938 Route Information w/ prefix smaller than this variable shall
1943 * 0 if accept_ra_rtr_pref is enabled.
1944 * -1 if accept_ra_rtr_pref is disabled.
1946 accept_ra_rt_info_max_plen - INTEGER
1947 Maximum prefix length of Route Information in RA.
1949 Route Information w/ prefix larger than this variable shall
1954 * 0 if accept_ra_rtr_pref is enabled.
1955 * -1 if accept_ra_rtr_pref is disabled.
1957 accept_ra_rtr_pref - BOOLEAN
1958 Accept Router Preference in RA.
1962 - enabled if accept_ra is enabled.
1963 - disabled if accept_ra is disabled.
1965 accept_ra_mtu - BOOLEAN
1966 Apply the MTU value specified in RA option 5 (RFC4861). If
1967 disabled, the MTU specified in the RA will be ignored.
1971 - enabled if accept_ra is enabled.
1972 - disabled if accept_ra is disabled.
1974 accept_redirects - BOOLEAN
1979 - enabled if local forwarding is disabled.
1980 - disabled if local forwarding is enabled.
1982 accept_source_route - INTEGER
1983 Accept source routing (routing extension header).
1985 - >= 0: Accept only routing header type 2.
1986 - < 0: Do not accept routing header.
1991 Autoconfigure addresses using Prefix Information in Router
1996 - enabled if accept_ra_pinfo is enabled.
1997 - disabled if accept_ra_pinfo is disabled.
1999 dad_transmits - INTEGER
2000 The amount of Duplicate Address Detection probes to send.
2004 forwarding - INTEGER
2005 Configure interface-specific Host/Router behaviour.
2009 It is recommended to have the same setting on all
2010 interfaces; mixed router/host scenarios are rather uncommon.
2012 Possible values are:
2014 - 0 Forwarding disabled
2015 - 1 Forwarding enabled
2019 By default, Host behaviour is assumed. This means:
2021 1. IsRouter flag is not set in Neighbour Advertisements.
2022 2. If accept_ra is TRUE (default), transmit Router
2024 3. If accept_ra is TRUE (default), accept Router
2025 Advertisements (and do autoconfiguration).
2026 4. If accept_redirects is TRUE (default), accept Redirects.
2030 If local forwarding is enabled, Router behaviour is assumed.
2031 This means exactly the reverse from the above:
2033 1. IsRouter flag is set in Neighbour Advertisements.
2034 2. Router Solicitations are not sent unless accept_ra is 2.
2035 3. Router Advertisements are ignored unless accept_ra is 2.
2036 4. Redirects are ignored.
2038 Default: 0 (disabled) if global forwarding is disabled (default),
2039 otherwise 1 (enabled).
2042 Default Hop Limit to set.
2047 Default Maximum Transfer Unit
2049 Default: 1280 (IPv6 required minimum)
2051 ip_nonlocal_bind - BOOLEAN
2052 If set, allows processes to bind() to non-local IPv6 addresses,
2053 which can be quite useful - but may break some applications.
2057 router_probe_interval - INTEGER
2058 Minimum interval (in seconds) between Router Probing described
2063 router_solicitation_delay - INTEGER
2064 Number of seconds to wait after interface is brought up
2065 before sending Router Solicitations.
2069 router_solicitation_interval - INTEGER
2070 Number of seconds to wait between Router Solicitations.
2074 router_solicitations - INTEGER
2075 Number of Router Solicitations to send until assuming no
2076 routers are present.
2080 use_oif_addrs_only - BOOLEAN
2081 When enabled, the candidate source addresses for destinations
2082 routed via this interface are restricted to the set of addresses
2083 configured on this interface (vis. RFC 6724, section 4).
2087 use_tempaddr - INTEGER
2088 Preference for Privacy Extensions (RFC3041).
2090 * <= 0 : disable Privacy Extensions
2091 * == 1 : enable Privacy Extensions, but prefer public
2092 addresses over temporary addresses.
2093 * > 1 : enable Privacy Extensions and prefer temporary
2094 addresses over public addresses.
2098 * 0 (for most devices)
2099 * -1 (for point-to-point devices and loopback devices)
2101 temp_valid_lft - INTEGER
2102 valid lifetime (in seconds) for temporary addresses.
2104 Default: 172800 (2 days)
2106 temp_prefered_lft - INTEGER
2107 Preferred lifetime (in seconds) for temporary addresses.
2109 Default: 86400 (1 day)
2111 keep_addr_on_down - INTEGER
2112 Keep all IPv6 addresses on an interface down event. If set static
2113 global addresses with no expiration time are not flushed.
2116 * 0 : system default
2119 Default: 0 (addresses are removed)
2121 max_desync_factor - INTEGER
2122 Maximum value for DESYNC_FACTOR, which is a random value
2123 that ensures that clients don't synchronize with each
2124 other and generate new addresses at exactly the same time.
2125 value is in seconds.
2129 regen_max_retry - INTEGER
2130 Number of attempts before give up attempting to generate
2131 valid temporary addresses.
2135 max_addresses - INTEGER
2136 Maximum number of autoconfigured addresses per interface. Setting
2137 to zero disables the limitation. It is not recommended to set this
2138 value too large (or to zero) because it would be an easy way to
2139 crash the kernel by allowing too many addresses to be created.
2143 disable_ipv6 - BOOLEAN
2144 Disable IPv6 operation. If accept_dad is set to 2, this value
2145 will be dynamically set to TRUE if DAD fails for the link-local
2148 Default: FALSE (enable IPv6 operation)
2150 When this value is changed from 1 to 0 (IPv6 is being enabled),
2151 it will dynamically create a link-local address on the given
2152 interface and start Duplicate Address Detection, if necessary.
2154 When this value is changed from 0 to 1 (IPv6 is being disabled),
2155 it will dynamically delete all addresses and routes on the given
2156 interface. From now on it will not possible to add addresses/routes
2157 to the selected interface.
2159 accept_dad - INTEGER
2160 Whether to accept DAD (Duplicate Address Detection).
2162 == ==============================================================
2164 1 Enable DAD (default)
2165 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate
2166 link-local address has been found.
2167 == ==============================================================
2169 DAD operation and mode on a given interface will be selected according
2170 to the maximum value of conf/{all,interface}/accept_dad.
2172 force_tllao - BOOLEAN
2173 Enable sending the target link-layer address option even when
2174 responding to a unicast neighbor solicitation.
2178 Quoting from RFC 2461, section 4.4, Target link-layer address:
2180 "The option MUST be included for multicast solicitations in order to
2181 avoid infinite Neighbor Solicitation "recursion" when the peer node
2182 does not have a cache entry to return a Neighbor Advertisements
2183 message. When responding to unicast solicitations, the option can be
2184 omitted since the sender of the solicitation has the correct link-
2185 layer address; otherwise it would not have be able to send the unicast
2186 solicitation in the first place. However, including the link-layer
2187 address in this case adds little overhead and eliminates a potential
2188 race condition where the sender deletes the cached link-layer address
2189 prior to receiving a response to a previous solicitation."
2191 ndisc_notify - BOOLEAN
2192 Define mode for notification of address and device changes.
2194 * 0 - (default): do nothing
2195 * 1 - Generate unsolicited neighbour advertisements when device is brought
2196 up or hardware address changes.
2198 ndisc_tclass - INTEGER
2199 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
2200 Discovery (Router Solicitation, Router Advertisement, Neighbor
2201 Solicitation, Neighbor Advertisement, Redirect) messages.
2202 These 8 bits can be interpreted as 6 high order bits holding the DSCP
2203 value and 2 low order bits representing ECN (which you probably want
2208 mldv1_unsolicited_report_interval - INTEGER
2209 The interval in milliseconds in which the next unsolicited
2210 MLDv1 report retransmit will take place.
2212 Default: 10000 (10 seconds)
2214 mldv2_unsolicited_report_interval - INTEGER
2215 The interval in milliseconds in which the next unsolicited
2216 MLDv2 report retransmit will take place.
2218 Default: 1000 (1 second)
2220 force_mld_version - INTEGER
2221 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
2222 * 1 - Enforce to use MLD version 1
2223 * 2 - Enforce to use MLD version 2
2225 suppress_frag_ndisc - INTEGER
2226 Control RFC 6980 (Security Implications of IPv6 Fragmentation
2227 with IPv6 Neighbor Discovery) behavior:
2229 * 1 - (default) discard fragmented neighbor discovery packets
2230 * 0 - allow fragmented neighbor discovery packets
2232 optimistic_dad - BOOLEAN
2233 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
2235 * 0: disabled (default)
2238 Optimistic Duplicate Address Detection for the interface will be enabled
2239 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
2240 it will be disabled otherwise.
2242 use_optimistic - BOOLEAN
2243 If enabled, do not classify optimistic addresses as deprecated during
2244 source address selection. Preferred addresses will still be chosen
2245 before optimistic addresses, subject to other ranking in the source
2246 address selection algorithm.
2248 * 0: disabled (default)
2251 This will be enabled if at least one of
2252 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
2254 stable_secret - IPv6 address
2255 This IPv6 address will be used as a secret to generate IPv6
2256 addresses for link-local addresses and autoconfigured
2257 ones. All addresses generated after setting this secret will
2258 be stable privacy ones by default. This can be changed via the
2259 addrgenmode ip-link. conf/default/stable_secret is used as the
2260 secret for the namespace, the interface specific ones can
2261 overwrite that. Writes to conf/all/stable_secret are refused.
2263 It is recommended to generate this secret during installation
2264 of a system and keep it stable after that.
2266 By default the stable secret is unset.
2268 addr_gen_mode - INTEGER
2269 Defines how link-local and autoconf addresses are generated.
2271 = =================================================================
2272 0 generate address based on EUI64 (default)
2273 1 do no generate a link-local address, use EUI64 for addresses
2274 generated from autoconf
2275 2 generate stable privacy addresses, using the secret from
2276 stable_secret (RFC7217)
2277 3 generate stable privacy addresses, using a random secret if unset
2278 = =================================================================
2280 drop_unicast_in_l2_multicast - BOOLEAN
2281 Drop any unicast IPv6 packets that are received in link-layer
2282 multicast (or broadcast) frames.
2284 By default this is turned off.
2286 drop_unsolicited_na - BOOLEAN
2287 Drop all unsolicited neighbor advertisements, for example if there's
2288 a known good NA proxy on the network and such frames need not be used
2289 (or in the case of 802.11, must not be used to prevent attacks.)
2291 By default this is turned off.
2293 enhanced_dad - BOOLEAN
2294 Include a nonce option in the IPv6 neighbor solicitation messages used for
2295 duplicate address detection per RFC7527. A received DAD NS will only signal
2296 a duplicate address if the nonce is different. This avoids any false
2297 detection of duplicates due to loopback of the NS messages that we send.
2298 The nonce option will be sent on an interface unless both of
2299 conf/{all,interface}/enhanced_dad are set to FALSE.
2307 Limit the maximal rates for sending ICMPv6 messages.
2309 0 to disable any limiting,
2310 otherwise the minimal space between responses in milliseconds.
2314 ratemask - list of comma separated ranges
2315 For ICMPv6 message types matching the ranges in the ratemask, limit
2316 the sending of the message according to ratelimit parameter.
2318 The format used for both input and output is a comma separated
2319 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
2320 129). Writing to the file will clear all previous ranges of ICMPv6
2321 message types and update the current list with the input.
2323 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
2324 for numerical values of ICMPv6 message types, e.g. echo request is 128
2325 and echo reply is 129.
2327 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
2329 echo_ignore_all - BOOLEAN
2330 If set non-zero, then the kernel will ignore all ICMP ECHO
2331 requests sent to it over the IPv6 protocol.
2335 echo_ignore_multicast - BOOLEAN
2336 If set non-zero, then the kernel will ignore all ICMP ECHO
2337 requests sent to it over the IPv6 protocol via multicast.
2341 echo_ignore_anycast - BOOLEAN
2342 If set non-zero, then the kernel will ignore all ICMP ECHO
2343 requests sent to it over the IPv6 protocol destined to anycast address.
2347 xfrm6_gc_thresh - INTEGER
2348 (Obsolete since linux-4.14)
2349 The threshold at which we will start garbage collecting for IPv6
2350 destination cache entries. At twice this value the system will
2351 refuse new allocations.
2355 Pekka Savola <pekkas@netcore.fi>
2356 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2359 /proc/sys/net/bridge/* Variables:
2360 =================================
2362 bridge-nf-call-arptables - BOOLEAN
2363 - 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2368 bridge-nf-call-iptables - BOOLEAN
2369 - 1 : pass bridged IPv4 traffic to iptables' chains.
2374 bridge-nf-call-ip6tables - BOOLEAN
2375 - 1 : pass bridged IPv6 traffic to ip6tables' chains.
2380 bridge-nf-filter-vlan-tagged - BOOLEAN
2381 - 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2386 bridge-nf-filter-pppoe-tagged - BOOLEAN
2387 - 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2392 bridge-nf-pass-vlan-input-dev - BOOLEAN
2393 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2394 interface on the bridge and set the netfilter input device to the
2395 vlan. This allows use of e.g. "iptables -i br0.1" and makes the
2396 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no
2397 matching vlan interface is found, or this switch is off, the input
2398 device is set to the bridge interface.
2400 - 0: disable bridge netfilter vlan interface lookup.
2404 ``proc/sys/net/sctp/*`` Variables:
2405 ==================================
2407 addip_enable - BOOLEAN
2408 Enable or disable extension of Dynamic Address Reconfiguration
2409 (ADD-IP) functionality specified in RFC5061. This extension provides
2410 the ability to dynamically add and remove new addresses for the SCTP
2413 1: Enable extension.
2415 0: Disable extension.
2420 Enable or disable pf (pf is short for potentially failed) state. A value
2421 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2422 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2423 Since pf_retrans and path_max_retrans can be changed by userspace
2424 application, sometimes user expects to disable pf state by the value of
2425 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2426 or path_max_retrans is changed by the user application, this pf state is
2427 enabled. As such, it is necessary to add this to dynamically enable
2428 and disable pf state. See:
2429 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2439 Unset or enable/disable pf (pf is short for potentially failed) state
2440 exposure. Applications can control the exposure of the PF path state
2441 in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2442 sockopt. When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2443 SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2444 can be got via SCTP_GET_PEER_ADDR_INFO sockopt; When it's enabled,
2445 a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2446 SCTP_PF state and a SCTP_PF-state transport info can be got via
2447 SCTP_GET_PEER_ADDR_INFO sockopt; When it's diabled, no
2448 SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2449 trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2452 0: Unset pf state exposure, Compatible with old applications.
2454 1: Disable pf state exposure.
2456 2: Enable pf state exposure.
2460 addip_noauth_enable - BOOLEAN
2461 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2462 authentication to protect the operations of adding or removing new
2463 addresses. This requirement is mandated so that unauthorized hosts
2464 would not be able to hijack associations. However, older
2465 implementations may not have implemented this requirement while
2466 allowing the ADD-IP extension. For reasons of interoperability,
2467 we provide this variable to control the enforcement of the
2468 authentication requirement.
2470 == ===============================================================
2471 1 Allow ADD-IP extension to be used without authentication. This
2472 should only be set in a closed environment for interoperability
2473 with older implementations.
2475 0 Enforce the authentication requirement
2476 == ===============================================================
2480 auth_enable - BOOLEAN
2481 Enable or disable Authenticated Chunks extension. This extension
2482 provides the ability to send and receive authenticated chunks and is
2483 required for secure operation of Dynamic Address Reconfiguration
2486 - 1: Enable this extension.
2487 - 0: Disable this extension.
2491 prsctp_enable - BOOLEAN
2492 Enable or disable the Partial Reliability extension (RFC3758) which
2493 is used to notify peers that a given DATA should no longer be expected.
2495 - 1: Enable extension
2501 The limit of the number of new packets that can be initially sent. It
2502 controls how bursty the generated traffic can be.
2506 association_max_retrans - INTEGER
2507 Set the maximum number for retransmissions that an association can
2508 attempt deciding that the remote end is unreachable. If this value
2509 is exceeded, the association is terminated.
2513 max_init_retransmits - INTEGER
2514 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2515 that an association will attempt before declaring the destination
2516 unreachable and terminating.
2520 path_max_retrans - INTEGER
2521 The maximum number of retransmissions that will be attempted on a given
2522 path. Once this threshold is exceeded, the path is considered
2523 unreachable, and new traffic will use a different path when the
2524 association is multihomed.
2528 pf_retrans - INTEGER
2529 The number of retransmissions that will be attempted on a given path
2530 before traffic is redirected to an alternate transport (should one
2531 exist). Note this is distinct from path_max_retrans, as a path that
2532 passes the pf_retrans threshold can still be used. Its only
2533 deprioritized when a transmission path is selected by the stack. This
2534 setting is primarily used to enable fast failover mechanisms without
2535 having to reduce path_max_retrans to a very low value. See:
2536 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2537 for details. Note also that a value of pf_retrans > path_max_retrans
2538 disables this feature. Since both pf_retrans and path_max_retrans can
2539 be changed by userspace application, a variable pf_enable is used to
2544 ps_retrans - INTEGER
2545 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
2546 from section-5 "Primary Path Switchover" in rfc7829. The primary path
2547 will be changed to another active path when the path error counter on
2548 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
2549 to continue data transmission on a new working path even when the old
2550 primary destination address becomes active again". Note this feature
2551 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
2552 and its value can't be less than 'pf_retrans' when changing by sysctl.
2556 rto_initial - INTEGER
2557 The initial round trip timeout value in milliseconds that will be used
2558 in calculating round trip times. This is the initial time interval
2559 for retransmissions.
2564 The maximum value (in milliseconds) of the round trip timeout. This
2565 is the largest time interval that can elapse between retransmissions.
2570 The minimum value (in milliseconds) of the round trip timeout. This
2571 is the smallest time interval the can elapse between retransmissions.
2575 hb_interval - INTEGER
2576 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2577 are sent at the specified interval on idle paths to probe the state of
2578 a given path between 2 associations.
2582 sack_timeout - INTEGER
2583 The amount of time (in milliseconds) that the implementation will wait
2588 valid_cookie_life - INTEGER
2589 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2590 is used during association establishment.
2594 cookie_preserve_enable - BOOLEAN
2595 Enable or disable the ability to extend the lifetime of the SCTP cookie
2596 that is used during the establishment phase of SCTP association
2598 - 1: Enable cookie lifetime extension.
2603 cookie_hmac_alg - STRING
2604 Select the hmac algorithm used when generating the cookie value sent by
2605 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2612 Ability to assign md5 or sha1 as the selected alg is predicated on the
2613 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2614 CONFIG_CRYPTO_SHA1).
2616 Default: Dependent on configuration. MD5 if available, else SHA1 if
2617 available, else none.
2619 rcvbuf_policy - INTEGER
2620 Determines if the receive buffer is attributed to the socket or to
2621 association. SCTP supports the capability to create multiple
2622 associations on a single socket. When using this capability, it is
2623 possible that a single stalled association that's buffering a lot
2624 of data may block other associations from delivering their data by
2625 consuming all of the receive buffer space. To work around this,
2626 the rcvbuf_policy could be set to attribute the receiver buffer space
2627 to each association instead of the socket. This prevents the described
2630 - 1: rcvbuf space is per association
2631 - 0: rcvbuf space is per socket
2635 sndbuf_policy - INTEGER
2636 Similar to rcvbuf_policy above, this applies to send buffer space.
2638 - 1: Send buffer is tracked per association
2639 - 0: Send buffer is tracked per socket.
2643 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2644 Number of pages allowed for queueing by all SCTP sockets.
2646 min: Below this number of pages SCTP is not bothered about its
2647 memory appetite. When amount of memory allocated by SCTP exceeds
2648 this number, SCTP starts to moderate memory usage.
2650 pressure: This value was introduced to follow format of tcp_mem.
2652 max: Number of pages allowed for queueing by all SCTP sockets.
2654 Default is calculated at boot time from amount of available memory.
2656 sctp_rmem - vector of 3 INTEGERs: min, default, max
2657 Only the first value ("min") is used, "default" and "max" are
2660 min: Minimal size of receive buffer used by SCTP socket.
2661 It is guaranteed to each SCTP socket (but not association) even
2662 under moderate memory pressure.
2666 sctp_wmem - vector of 3 INTEGERs: min, default, max
2667 Only the first value ("min") is used, "default" and "max" are
2670 min: Minimum size of send buffer that can be used by SCTP sockets.
2671 It is guaranteed to each SCTP socket (but not association) even
2672 under moderate memory pressure.
2676 addr_scope_policy - INTEGER
2677 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2679 - 0 - Disable IPv4 address scoping
2680 - 1 - Enable IPv4 address scoping
2681 - 2 - Follow draft but allow IPv4 private addresses
2682 - 3 - Follow draft but allow IPv4 link local addresses
2687 ``/proc/sys/net/core/*``
2688 ========================
2690 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
2693 ``/proc/sys/net/unix/*``
2694 ========================
2696 max_dgram_qlen - INTEGER
2697 The maximum length of dgram socket receive queue