1 .. SPDX-License-Identifier: GPL-2.0
7 /proc/sys/net/ipv4/* Variables
8 ==============================
11 - 0 - disabled (default)
14 Forward Packets between interfaces.
16 This variable is special, its change resets all configuration
17 parameters to their default state (RFC1122 for hosts, RFC1812
20 ip_default_ttl - INTEGER
21 Default value of TTL field (Time To Live) for outgoing (but not
22 forwarded) IP packets. Should be between 1 and 255 inclusive.
23 Default: 64 (as recommended by RFC1700)
25 ip_no_pmtu_disc - INTEGER
26 Disable Path MTU Discovery. If enabled in mode 1 and a
27 fragmentation-required ICMP is received, the PMTU to this
28 destination will be set to the smallest of the old MTU to
29 this destination and min_pmtu (see below). You will need
30 to raise min_pmtu to the smallest interface MTU on your system
31 manually if you want to avoid locally generated fragments.
33 In mode 2 incoming Path MTU Discovery messages will be
34 discarded. Outgoing frames are handled the same as in mode 1,
35 implicitly setting IP_PMTUDISC_DONT on every created socket.
37 Mode 3 is a hardened pmtu discover mode. The kernel will only
38 accept fragmentation-needed errors if the underlying protocol
39 can verify them besides a plain socket lookup. Current
40 protocols for which pmtu events will be honored are TCP, SCTP
41 and DCCP as they verify e.g. the sequence number or the
42 association. This mode should not be enabled globally but is
43 only intended to secure e.g. name servers in namespaces where
44 TCP path mtu must still work but path MTU information of other
45 protocols should be discarded. If enabled globally this mode
46 could break other protocols.
53 default 552 - minimum Path MTU. Unless this is changed manually,
54 each cached pmtu will never be lower than this setting.
56 ip_forward_use_pmtu - BOOLEAN
57 By default we don't trust protocol path MTUs while forwarding
58 because they could be easily forged and can lead to unwanted
59 fragmentation by the router.
60 You only need to enable this if you have user-space software
61 which tries to discover path mtus by itself and depends on the
62 kernel honoring this information. This is normally not the
72 fwmark_reflect - BOOLEAN
73 Controls the fwmark of kernel-generated IPv4 reply packets that are not
74 associated with a socket for example, TCP RSTs or ICMP echo replies).
75 If unset, these packets have a fwmark of zero. If set, they have the
76 fwmark of the packet they are replying to.
80 fib_multipath_use_neigh - BOOLEAN
81 Use status of existing neighbor entry when determining nexthop for
82 multipath routes. If disabled, neighbor information is not used and
83 packets could be directed to a failed nexthop. Only valid for kernels
84 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
93 fib_multipath_hash_policy - INTEGER
94 Controls which hash policy to use for multipath routes. Only valid
95 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
103 - 2 - Layer 3 or inner Layer 3 if present
104 - 3 - Custom multipath hash. Fields used for multipath hash calculation
105 are determined by fib_multipath_hash_fields sysctl
107 fib_multipath_hash_fields - UNSIGNED INTEGER
108 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
109 fields used for multipath hash calculation are determined by this
112 This value is a bitmask which enables various fields for multipath hash
117 ====== ============================
118 0x0001 Source IP address
119 0x0002 Destination IP address
121 0x0008 Unused (Flow Label)
123 0x0020 Destination port
124 0x0040 Inner source IP address
125 0x0080 Inner destination IP address
126 0x0100 Inner IP protocol
127 0x0200 Inner Flow Label
128 0x0400 Inner source port
129 0x0800 Inner destination port
130 ====== ============================
132 Default: 0x0007 (source IP, destination IP and IP protocol)
134 fib_sync_mem - UNSIGNED INTEGER
135 Amount of dirty memory from fib entries that can be backlogged before
136 synchronize_rcu is forced.
138 Default: 512kB Minimum: 64kB Maximum: 64MB
140 ip_forward_update_priority - INTEGER
141 Whether to update SKB priority from "TOS" field in IPv4 header after it
142 is forwarded. The new SKB priority is mapped from TOS field value
143 according to an rt_tos2priority table (see e.g. man tc-prio).
145 Default: 1 (Update priority.)
149 - 0 - Do not update priority.
150 - 1 - Update priority.
152 route/max_size - INTEGER
153 Maximum number of routes allowed in the kernel. Increase
154 this when using large numbers of interfaces and/or routes.
156 From linux kernel 3.6 onwards, this is deprecated for ipv4
157 as route cache is no longer used.
159 From linux kernel 6.3 onwards, this is deprecated for ipv6
160 as garbage collection manages cached route entries.
162 neigh/default/gc_thresh1 - INTEGER
163 Minimum number of entries to keep. Garbage collector will not
164 purge entries if there are fewer than this number.
168 neigh/default/gc_thresh2 - INTEGER
169 Threshold when garbage collector becomes more aggressive about
170 purging entries. Entries older than 5 seconds will be cleared
171 when over this number.
175 neigh/default/gc_thresh3 - INTEGER
176 Maximum number of non-PERMANENT neighbor entries allowed. Increase
177 this when using large numbers of interfaces and when communicating
178 with large numbers of directly-connected peers.
182 neigh/default/unres_qlen_bytes - INTEGER
183 The maximum number of bytes which may be used by packets
184 queued for each unresolved address by other network layers.
187 Setting negative value is meaningless and will return error.
189 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
191 Exact value depends on architecture and kernel options,
192 but should be enough to allow queuing 256 packets
195 neigh/default/unres_qlen - INTEGER
196 The maximum number of packets which may be queued for each
197 unresolved address by other network layers.
199 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
201 Prior to linux 3.3, the default value is 3 which may cause
202 unexpected packet loss. The current default value is calculated
203 according to default value of unres_qlen_bytes and true size of
208 neigh/default/interval_probe_time_ms - INTEGER
209 The probe interval for neighbor entries with NTF_MANAGED flag,
214 mtu_expires - INTEGER
215 Time, in seconds, that cached PMTU information is kept.
217 min_adv_mss - INTEGER
218 The advertised MSS depends on the first hop route MTU, but will
219 never be lower than this setting.
221 fib_notify_on_flag_change - INTEGER
222 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
223 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
225 After installing a route to the kernel, user space receives an
226 acknowledgment, which means the route was installed in the kernel,
227 but not necessarily in hardware.
228 It is also possible for a route already installed in hardware to change
229 its action and therefore its flags. For example, a host route that is
230 trapping packets can be "promoted" to perform decapsulation following
231 the installation of an IPinIP/VXLAN tunnel.
232 The notifications will indicate to user-space the state of the route.
234 Default: 0 (Do not emit notifications.)
238 - 0 - Do not emit notifications.
239 - 1 - Emit notifications.
240 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
244 ipfrag_high_thresh - LONG INTEGER
245 Maximum memory used to reassemble IP fragments.
247 ipfrag_low_thresh - LONG INTEGER
248 (Obsolete since linux-4.17)
249 Maximum memory used to reassemble IP fragments before the kernel
250 begins to remove incomplete fragment queues to free up resources.
251 The kernel still accepts new fragments for defragmentation.
253 ipfrag_time - INTEGER
254 Time in seconds to keep an IP fragment in memory.
256 ipfrag_max_dist - INTEGER
257 ipfrag_max_dist is a non-negative integer value which defines the
258 maximum "disorder" which is allowed among fragments which share a
259 common IP source address. Note that reordering of packets is
260 not unusual, but if a large number of fragments arrive from a source
261 IP address while a particular fragment queue remains incomplete, it
262 probably indicates that one or more fragments belonging to that queue
263 have been lost. When ipfrag_max_dist is positive, an additional check
264 is done on fragments before they are added to a reassembly queue - if
265 ipfrag_max_dist (or more) fragments have arrived from a particular IP
266 address between additions to any IP fragment queue using that source
267 address, it's presumed that one or more fragments in the queue are
268 lost. The existing fragment queue will be dropped, and a new one
269 started. An ipfrag_max_dist value of zero disables this check.
271 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
272 result in unnecessarily dropping fragment queues when normal
273 reordering of packets occurs, which could lead to poor application
274 performance. Using a very large value, e.g. 50000, increases the
275 likelihood of incorrectly reassembling IP fragments that originate
276 from different IP datagrams, which could result in data corruption.
279 bc_forwarding - INTEGER
280 bc_forwarding enables the feature described in rfc1812#section-5.3.5.2
281 and rfc2644. It allows the router to forward directed broadcast.
282 To enable this feature, the 'all' entry and the input interface entry
289 inet_peer_threshold - INTEGER
290 The approximate size of the storage. Starting from this threshold
291 entries will be thrown aggressively. This threshold also determines
292 entries' time-to-live and time intervals between garbage collection
293 passes. More entries, less time-to-live, less GC interval.
295 inet_peer_minttl - INTEGER
296 Minimum time-to-live of entries. Should be enough to cover fragment
297 time-to-live on the reassembling side. This minimum time-to-live is
298 guaranteed if the pool size is less than inet_peer_threshold.
301 inet_peer_maxttl - INTEGER
302 Maximum time-to-live of entries. Unused entries will expire after
303 this period of time if there is no memory pressure on the pool (i.e.
304 when the number of entries in the pool is very small).
311 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
312 Defaults to 4096. (Was 128 before linux-5.4)
313 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
315 tcp_abort_on_overflow - BOOLEAN
316 If listening service is too slow to accept new connections,
317 reset them. Default state is FALSE. It means that if overflow
318 occurred due to a burst, connection will recover. Enable this
319 option _only_ if you are really sure that listening daemon
320 cannot be tuned to accept connections faster. Enabling this
321 option can harm clients of your server.
323 tcp_adv_win_scale - INTEGER
324 Count buffering overhead as bytes/2^tcp_adv_win_scale
325 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
328 Possible values are [-31, 31], inclusive.
332 tcp_allowed_congestion_control - STRING
333 Show/set the congestion control choices available to non-privileged
334 processes. The list is a subset of those listed in
335 tcp_available_congestion_control.
337 Default is "reno" and the default setting (tcp_congestion_control).
339 tcp_app_win - INTEGER
340 Reserve max(window/2^tcp_app_win, mss) of window for application
341 buffer. Value 0 is special, it means that nothing is reserved.
343 Possible values are [0, 31], inclusive.
347 tcp_autocorking - BOOLEAN
348 Enable TCP auto corking :
349 When applications do consecutive small write()/sendmsg() system calls,
350 we try to coalesce these small writes as much as possible, to lower
351 total amount of sent packets. This is done if at least one prior
352 packet for the flow is waiting in Qdisc queues or device transmit
353 queue. Applications can still use TCP_CORK for optimal behavior
354 when they know how/when to uncork their sockets.
358 tcp_available_congestion_control - STRING
359 Shows the available congestion control choices that are registered.
360 More congestion control algorithms may be available as modules,
363 tcp_base_mss - INTEGER
364 The initial value of search_low to be used by the packetization layer
365 Path MTU discovery (MTU probing). If MTU probing is enabled,
366 this is the initial MSS used by the connection.
368 tcp_mtu_probe_floor - INTEGER
369 If MTU probing is enabled this caps the minimum MSS used for search_low
374 tcp_min_snd_mss - INTEGER
375 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
376 as described in RFC 1122 and RFC 6691.
378 If this ADVMSS option is smaller than tcp_min_snd_mss,
379 it is silently capped to tcp_min_snd_mss.
381 Default : 48 (at least 8 bytes of payload per segment)
383 tcp_congestion_control - STRING
384 Set the congestion control algorithm to be used for new
385 connections. The algorithm "reno" is always available, but
386 additional choices may be available based on kernel configuration.
387 Default is set as part of kernel configuration.
388 For passive connections, the listener congestion control choice
391 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
394 Allows TCP to send "duplicate" SACKs.
396 tcp_early_retrans - INTEGER
397 Tail loss probe (TLP) converts RTOs occurring due to tail
398 losses into fast recovery (draft-ietf-tcpm-rack). Note that
399 TLP requires RACK to function properly (see tcp_recovery below)
409 Control use of Explicit Congestion Notification (ECN) by TCP.
410 ECN is used only when both ends of the TCP connection indicate
411 support for it. This feature is useful in avoiding losses due
412 to congestion by allowing supporting routers to signal
413 congestion before having to drop packets.
417 = =====================================================
418 0 Disable ECN. Neither initiate nor accept ECN.
419 1 Enable ECN when requested by incoming connections and
420 also request ECN on outgoing connection attempts.
421 2 Enable ECN when requested by incoming connections
422 but do not request ECN on outgoing connections.
423 = =====================================================
427 tcp_ecn_fallback - BOOLEAN
428 If the kernel detects that ECN connection misbehaves, enable fall
429 back to non-ECN. Currently, this knob implements the fallback
430 from RFC3168, section 6.1.1.1., but we reserve that in future,
431 additional detection mechanisms could be implemented under this
432 knob. The value is not used, if tcp_ecn or per route (or congestion
433 control) ECN settings are disabled.
435 Default: 1 (fallback enabled)
438 This is a legacy option, it has no effect anymore.
440 tcp_fin_timeout - INTEGER
441 The length of time an orphaned (no longer referenced by any
442 application) connection will remain in the FIN_WAIT_2 state
443 before it is aborted at the local end. While a perfectly
444 valid "receive only" state for an un-orphaned connection, an
445 orphaned connection in FIN_WAIT_2 state could otherwise wait
446 forever for the remote to close its end of the connection.
453 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
454 F-RTO is an enhanced recovery algorithm for TCP retransmission
455 timeouts. It is particularly beneficial in networks where the
456 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
457 modification. It does not require any support from the peer.
459 By default it's enabled with a non-zero value. 0 disables F-RTO.
461 tcp_fwmark_accept - BOOLEAN
462 If set, incoming connections to listening sockets that do not have a
463 socket mark will set the mark of the accepting socket to the fwmark of
464 the incoming SYN packet. This will cause all packets on that connection
465 (starting from the first SYNACK) to be sent with that fwmark. The
466 listening socket's mark is unchanged. Listening sockets that already
467 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
472 tcp_invalid_ratelimit - INTEGER
473 Limit the maximal rate for sending duplicate acknowledgments
474 in response to incoming TCP packets that are for an existing
475 connection but that are invalid due to any of these reasons:
477 (a) out-of-window sequence number,
478 (b) out-of-window acknowledgment number, or
479 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
481 This can help mitigate simple "ack loop" DoS attacks, wherein
482 a buggy or malicious middlebox or man-in-the-middle can
483 rewrite TCP header fields in manner that causes each endpoint
484 to think that the other is sending invalid TCP segments, thus
485 causing each side to send an unterminating stream of duplicate
486 acknowledgments for invalid segments.
488 Using 0 disables rate-limiting of dupacks in response to
489 invalid segments; otherwise this value specifies the minimal
490 space between sending such dupacks, in milliseconds.
492 Default: 500 (milliseconds).
494 tcp_keepalive_time - INTEGER
495 How often TCP sends out keepalive messages when keepalive is enabled.
498 tcp_keepalive_probes - INTEGER
499 How many keepalive probes TCP sends out, until it decides that the
500 connection is broken. Default value: 9.
502 tcp_keepalive_intvl - INTEGER
503 How frequently the probes are send out. Multiplied by
504 tcp_keepalive_probes it is time to kill not responding connection,
505 after probes started. Default value: 75sec i.e. connection
506 will be aborted after ~11 minutes of retries.
508 tcp_l3mdev_accept - BOOLEAN
509 Enables child sockets to inherit the L3 master device index.
510 Enabling this option allows a "global" listen socket to work
511 across L3 master domains (e.g., VRFs) with connected sockets
512 derived from the listen socket to be bound to the L3 domain in
513 which the packets originated. Only valid when the kernel was
514 compiled with CONFIG_NET_L3_MASTER_DEV.
516 Default: 0 (disabled)
518 tcp_low_latency - BOOLEAN
519 This is a legacy option, it has no effect anymore.
521 tcp_max_orphans - INTEGER
522 Maximal number of TCP sockets not attached to any user file handle,
523 held by system. If this number is exceeded orphaned connections are
524 reset immediately and warning is printed. This limit exists
525 only to prevent simple DoS attacks, you _must_ not rely on this
526 or lower the limit artificially, but rather increase it
527 (probably, after increasing installed memory),
528 if network conditions require more than default value,
529 and tune network services to linger and kill such states
530 more aggressively. Let me to remind again: each orphan eats
531 up to ~64K of unswappable memory.
533 tcp_max_syn_backlog - INTEGER
534 Maximal number of remembered connection requests (SYN_RECV),
535 which have not received an acknowledgment from connecting client.
537 This is a per-listener limit.
539 The minimal value is 128 for low memory machines, and it will
540 increase in proportion to the memory of machine.
542 If server suffers from overload, try increasing this number.
544 Remember to also check /proc/sys/net/core/somaxconn
545 A SYN_RECV request socket consumes about 304 bytes of memory.
547 tcp_max_tw_buckets - INTEGER
548 Maximal number of timewait sockets held by system simultaneously.
549 If this number is exceeded time-wait socket is immediately destroyed
550 and warning is printed. This limit exists only to prevent
551 simple DoS attacks, you _must_ not lower the limit artificially,
552 but rather increase it (probably, after increasing installed memory),
553 if network conditions require more than default value.
555 tcp_mem - vector of 3 INTEGERs: min, pressure, max
556 min: below this number of pages TCP is not bothered about its
559 pressure: when amount of memory allocated by TCP exceeds this number
560 of pages, TCP moderates its memory consumption and enters memory
561 pressure mode, which is exited when memory consumption falls
564 max: number of pages allowed for queueing by all TCP sockets.
566 Defaults are calculated at boot time from amount of available
569 tcp_min_rtt_wlen - INTEGER
570 The window length of the windowed min filter to track the minimum RTT.
571 A shorter window lets a flow more quickly pick up new (higher)
572 minimum RTT when it is moved to a longer path (e.g., due to traffic
573 engineering). A longer window makes the filter more resistant to RTT
574 inflations such as transient congestion. The unit is seconds.
576 Possible values: 0 - 86400 (1 day)
580 tcp_moderate_rcvbuf - BOOLEAN
581 If set, TCP performs receive buffer auto-tuning, attempting to
582 automatically size the buffer (no greater than tcp_rmem[2]) to
583 match the size required by the path for full throughput. Enabled by
586 tcp_mtu_probing - INTEGER
587 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
591 - 1 - Disabled by default, enabled when an ICMP black hole detected
592 - 2 - Always enabled, use initial MSS of tcp_base_mss.
594 tcp_probe_interval - UNSIGNED INTEGER
595 Controls how often to start TCP Packetization-Layer Path MTU
596 Discovery reprobe. The default is reprobing every 10 minutes as
599 tcp_probe_threshold - INTEGER
600 Controls when TCP Packetization-Layer Path MTU Discovery probing
601 will stop in respect to the width of search range in bytes. Default
604 tcp_no_metrics_save - BOOLEAN
605 By default, TCP saves various connection metrics in the route cache
606 when the connection closes, so that connections established in the
607 near future can use these to set initial conditions. Usually, this
608 increases overall performance, but may sometimes cause performance
609 degradation. If set, TCP will not cache metrics on closing
612 tcp_no_ssthresh_metrics_save - BOOLEAN
613 Controls whether TCP saves ssthresh metrics in the route cache.
615 Default is 1, which disables ssthresh metrics.
617 tcp_orphan_retries - INTEGER
618 This value influences the timeout of a locally closed TCP connection,
619 when RTO retransmissions remain unacknowledged.
620 See tcp_retries2 for more details.
622 The default value is 8.
624 If your machine is a loaded WEB server,
625 you should think about lowering this value, such sockets
626 may consume significant resources. Cf. tcp_max_orphans.
628 tcp_recovery - INTEGER
629 This value is a bitmap to enable various experimental loss recovery
632 ========= =============================================================
633 RACK: 0x1 enables the RACK loss detection for fast detection of lost
634 retransmissions and tail drops. It also subsumes and disables
635 RFC6675 recovery for SACK connections.
637 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
639 RACK: 0x4 disables RACK's DUPACK threshold heuristic
640 ========= =============================================================
644 tcp_reflect_tos - BOOLEAN
645 For listening sockets, reuse the DSCP value of the initial SYN message
646 for outgoing packets. This allows to have both directions of a TCP
647 stream to use the same DSCP value, assuming DSCP remains unchanged for
648 the lifetime of the connection.
650 This options affects both IPv4 and IPv6.
652 Default: 0 (disabled)
654 tcp_reordering - INTEGER
655 Initial reordering level of packets in a TCP stream.
656 TCP stack can then dynamically adjust flow reordering level
657 between this initial value and tcp_max_reordering
661 tcp_max_reordering - INTEGER
662 Maximal reordering level of packets in a TCP stream.
663 300 is a fairly conservative value, but you might increase it
664 if paths are using per packet load balancing (like bonding rr mode)
668 tcp_retrans_collapse - BOOLEAN
669 Bug-to-bug compatibility with some broken printers.
670 On retransmit try to send bigger packets to work around bugs in
673 tcp_retries1 - INTEGER
674 This value influences the time, after which TCP decides, that
675 something is wrong due to unacknowledged RTO retransmissions,
676 and reports this suspicion to the network layer.
677 See tcp_retries2 for more details.
679 RFC 1122 recommends at least 3 retransmissions, which is the
682 tcp_retries2 - INTEGER
683 This value influences the timeout of an alive TCP connection,
684 when RTO retransmissions remain unacknowledged.
685 Given a value of N, a hypothetical TCP connection following
686 exponential backoff with an initial RTO of TCP_RTO_MIN would
687 retransmit N times before killing the connection at the (N+1)th RTO.
689 The default value of 15 yields a hypothetical timeout of 924.6
690 seconds and is a lower bound for the effective timeout.
691 TCP will effectively time out at the first RTO which exceeds the
692 hypothetical timeout.
694 RFC 1122 recommends at least 100 seconds for the timeout,
695 which corresponds to a value of at least 8.
697 tcp_rfc1337 - BOOLEAN
698 If set, the TCP stack behaves conforming to RFC1337. If unset,
699 we are not conforming to RFC, but prevent TCP TIME_WAIT
704 tcp_rmem - vector of 3 INTEGERs: min, default, max
705 min: Minimal size of receive buffer used by TCP sockets.
706 It is guaranteed to each TCP socket, even under moderate memory
711 default: initial size of receive buffer used by TCP sockets.
712 This value overrides net.core.rmem_default used by other protocols.
713 Default: 131072 bytes.
714 This value results in initial window of 65535.
716 max: maximal size of receive buffer allowed for automatically
717 selected receiver buffers for TCP socket. This value does not override
718 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
719 automatic tuning of that socket's receive buffer size, in which
720 case this value is ignored.
721 Default: between 131072 and 6MB, depending on RAM size.
724 Enable select acknowledgments (SACKS).
726 tcp_comp_sack_delay_ns - LONG INTEGER
727 TCP tries to reduce number of SACK sent, using a timer
728 based on 5% of SRTT, capped by this sysctl, in nano seconds.
729 The default is 1ms, based on TSO autosizing period.
731 Default : 1,000,000 ns (1 ms)
733 tcp_comp_sack_slack_ns - LONG INTEGER
734 This sysctl control the slack used when arming the
735 timer used by SACK compression. This gives extra time
736 for small RTT flows, and reduces system overhead by allowing
737 opportunistic reduction of timer interrupts.
739 Default : 100,000 ns (100 us)
741 tcp_comp_sack_nr - INTEGER
742 Max number of SACK that can be compressed.
743 Using 0 disables SACK compression.
747 tcp_slow_start_after_idle - BOOLEAN
748 If set, provide RFC2861 behavior and time out the congestion
749 window after an idle period. An idle period is defined at
750 the current RTO. If unset, the congestion window will not
751 be timed out after an idle period.
756 Use the Host requirements interpretation of the TCP urgent pointer field.
757 Most hosts use the older BSD interpretation, so if you turn this on
758 Linux might not communicate correctly with them.
762 tcp_synack_retries - INTEGER
763 Number of times SYNACKs for a passive TCP connection attempt will
764 be retransmitted. Should not be higher than 255. Default value
765 is 5, which corresponds to 31seconds till the last retransmission
766 with the current initial RTO of 1second. With this the final timeout
767 for a passive TCP connection will happen after 63seconds.
769 tcp_syncookies - INTEGER
770 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
771 Send out syncookies when the syn backlog queue of a socket
772 overflows. This is to prevent against the common 'SYN flood attack'
775 Note, that syncookies is fallback facility.
776 It MUST NOT be used to help highly loaded servers to stand
777 against legal connection rate. If you see SYN flood warnings
778 in your logs, but investigation shows that they occur
779 because of overload with legal connections, you should tune
780 another parameters until this warning disappear.
781 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
783 syncookies seriously violate TCP protocol, do not allow
784 to use TCP extensions, can result in serious degradation
785 of some services (f.e. SMTP relaying), visible not by you,
786 but your clients and relays, contacting you. While you see
787 SYN flood warnings in logs not being really flooded, your server
788 is seriously misconfigured.
790 If you want to test which effects syncookies have to your
791 network connections you can set this knob to 2 to enable
792 unconditionally generation of syncookies.
794 tcp_migrate_req - BOOLEAN
795 The incoming connection is tied to a specific listening socket when
796 the initial SYN packet is received during the three-way handshake.
797 When a listener is closed, in-flight request sockets during the
798 handshake and established sockets in the accept queue are aborted.
800 If the listener has SO_REUSEPORT enabled, other listeners on the
801 same port should have been able to accept such connections. This
802 option makes it possible to migrate such child sockets to another
803 listener after close() or shutdown().
805 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
806 usually be used to define the policy to pick an alive listener.
807 Otherwise, the kernel will randomly pick an alive listener only if
808 this option is enabled.
810 Note that migration between listeners with different settings may
811 crash applications. Let's say migration happens from listener A to
812 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
813 the requests migrated from A. To avoid such a situation, cancel
814 migration by returning SK_DROP in the type of eBPF program, or
819 tcp_fastopen - INTEGER
820 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
823 The client support is enabled by flag 0x1 (on by default). The client
824 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
825 rather than connect() to send data in SYN.
827 The server support is enabled by flag 0x2 (off by default). Then
828 either enable for all listeners with another flag (0x400) or
829 enable individual listeners via TCP_FASTOPEN socket option with
830 the option value being the length of the syn-data backlog.
832 The values (bitmap) are
834 ===== ======== ======================================================
835 0x1 (client) enables sending data in the opening SYN on the client.
836 0x2 (server) enables the server support, i.e., allowing data in
837 a SYN packet to be accepted and passed to the
838 application before 3-way handshake finishes.
839 0x4 (client) send data in the opening SYN regardless of cookie
840 availability and without a cookie option.
841 0x200 (server) accept data-in-SYN w/o any cookie option present.
842 0x400 (server) enable all listeners to support Fast Open by
843 default without explicit TCP_FASTOPEN socket option.
844 ===== ======== ======================================================
848 Note that additional client or server features are only
849 effective if the basic support (0x1 and 0x2) are enabled respectively.
851 tcp_fastopen_blackhole_timeout_sec - INTEGER
852 Initial time period in second to disable Fastopen on active TCP sockets
853 when a TFO firewall blackhole issue happens.
854 This time period will grow exponentially when more blackhole issues
855 get detected right after Fastopen is re-enabled and will reset to
856 initial value when the blackhole issue goes away.
857 0 to disable the blackhole detection.
859 By default, it is set to 0 (feature is disabled).
861 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
862 The list consists of a primary key and an optional backup key. The
863 primary key is used for both creating and validating cookies, while the
864 optional backup key is only used for validating cookies. The purpose of
865 the backup key is to maximize TFO validation when keys are rotated.
867 A randomly chosen primary key may be configured by the kernel if
868 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
869 TCP_FASTOPEN setsockopt() optname is set and a key has not been
870 previously configured via sysctl. If keys are configured via
871 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
872 per-socket keys will be used instead of any keys that are specified via
875 A key is specified as 4 8-digit hexadecimal integers which are separated
876 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
877 omitted. A primary and a backup key may be specified by separating them
878 by a comma. If only one key is specified, it becomes the primary key and
879 any previously configured backup keys are removed.
881 tcp_syn_retries - INTEGER
882 Number of times initial SYNs for an active TCP connection attempt
883 will be retransmitted. Should not be higher than 127. Default value
884 is 6, which corresponds to 67seconds (with tcp_syn_linear_timeouts = 4)
885 till the last retransmission with the current initial RTO of 1second.
886 With this the final timeout for an active TCP connection attempt
887 will happen after 131seconds.
889 tcp_timestamps - INTEGER
890 Enable timestamps as defined in RFC1323.
893 - 1: Enable timestamps as defined in RFC1323 and use random offset for
894 each connection rather than only using the current time.
895 - 2: Like 1, but without random offsets.
899 tcp_min_tso_segs - INTEGER
900 Minimal number of segments per TSO frame.
902 Since linux-3.12, TCP does an automatic sizing of TSO frames,
903 depending on flow rate, instead of filling 64Kbytes packets.
904 For specific usages, it's possible to force TCP to build big
905 TSO frames. Note that TCP stack might split too big TSO packets
906 if available window is too small.
910 tcp_tso_rtt_log - INTEGER
911 Adjustment of TSO packet sizes based on min_rtt
913 Starting from linux-5.18, TCP autosizing can be tweaked
914 for flows having small RTT.
916 Old autosizing was splitting the pacing budget to send 1024 TSO
919 tso_packet_size = sk->sk_pacing_rate / 1024;
921 With the new mechanism, we increase this TSO sizing using:
923 distance = min_rtt_usec / (2^tcp_tso_rtt_log)
924 tso_packet_size += gso_max_size >> distance;
926 This means that flows between very close hosts can use bigger
927 TSO packets, reducing their cpu costs.
929 If you want to use the old autosizing, set this sysctl to 0.
931 Default: 9 (2^9 = 512 usec)
933 tcp_pacing_ss_ratio - INTEGER
934 sk->sk_pacing_rate is set by TCP stack using a ratio applied
935 to current rate. (current_rate = cwnd * mss / srtt)
936 If TCP is in slow start, tcp_pacing_ss_ratio is applied
937 to let TCP probe for bigger speeds, assuming cwnd can be
938 doubled every other RTT.
942 tcp_pacing_ca_ratio - INTEGER
943 sk->sk_pacing_rate is set by TCP stack using a ratio applied
944 to current rate. (current_rate = cwnd * mss / srtt)
945 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
946 is applied to conservatively probe for bigger throughput.
950 tcp_syn_linear_timeouts - INTEGER
951 The number of times for an active TCP connection to retransmit SYNs with
952 a linear backoff timeout before defaulting to an exponential backoff
953 timeout. This has no effect on SYNACK at the passive TCP side.
955 With an initial RTO of 1 and tcp_syn_linear_timeouts = 4 we would
956 expect SYN RTOs to be: 1, 1, 1, 1, 1, 2, 4, ... (4 linear timeouts,
957 and the first exponential backoff using 2^0 * initial_RTO).
960 tcp_tso_win_divisor - INTEGER
961 This allows control over what percentage of the congestion window
962 can be consumed by a single TSO frame.
963 The setting of this parameter is a choice between burstiness and
964 building larger TSO frames.
968 tcp_tw_reuse - INTEGER
969 Enable reuse of TIME-WAIT sockets for new connections when it is
970 safe from protocol viewpoint.
974 - 2 - enable for loopback traffic only
976 It should not be changed without advice/request of technical
981 tcp_window_scaling - BOOLEAN
982 Enable window scaling as defined in RFC1323.
984 tcp_shrink_window - BOOLEAN
985 This changes how the TCP receive window is calculated.
987 RFC 7323, section 2.4, says there are instances when a retracted
988 window can be offered, and that TCP implementations MUST ensure
989 that they handle a shrinking window, as specified in RFC 1122.
991 - 0 - Disabled. The window is never shrunk.
992 - 1 - Enabled. The window is shrunk when necessary to remain within
993 the memory limit set by autotuning (sk_rcvbuf).
994 This only occurs if a non-zero receive window
995 scaling factor is also in effect.
999 tcp_wmem - vector of 3 INTEGERs: min, default, max
1000 min: Amount of memory reserved for send buffers for TCP sockets.
1001 Each TCP socket has rights to use it due to fact of its birth.
1005 default: initial size of send buffer used by TCP sockets. This
1006 value overrides net.core.wmem_default used by other protocols.
1008 It is usually lower than net.core.wmem_default.
1012 max: Maximal amount of memory allowed for automatically tuned
1013 send buffers for TCP sockets. This value does not override
1014 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
1015 automatic tuning of that socket's send buffer size, in which case
1016 this value is ignored.
1018 Default: between 64K and 4MB, depending on RAM size.
1020 tcp_notsent_lowat - UNSIGNED INTEGER
1021 A TCP socket can control the amount of unsent bytes in its write queue,
1022 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
1023 reports POLLOUT events if the amount of unsent bytes is below a per
1024 socket value, and if the write queue is not full. sendmsg() will
1025 also not add new buffers if the limit is hit.
1027 This global variable controls the amount of unsent data for
1028 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
1029 to the global variable has immediate effect.
1031 Default: UINT_MAX (0xFFFFFFFF)
1033 tcp_workaround_signed_windows - BOOLEAN
1034 If set, assume no receipt of a window scaling option means the
1035 remote TCP is broken and treats the window as a signed quantity.
1036 If unset, assume the remote TCP is not broken even if we do
1037 not receive a window scaling option from them.
1041 tcp_thin_linear_timeouts - BOOLEAN
1042 Enable dynamic triggering of linear timeouts for thin streams.
1043 If set, a check is performed upon retransmission by timeout to
1044 determine if the stream is thin (less than 4 packets in flight).
1045 As long as the stream is found to be thin, up to 6 linear
1046 timeouts may be performed before exponential backoff mode is
1047 initiated. This improves retransmission latency for
1048 non-aggressive thin streams, often found to be time-dependent.
1049 For more information on thin streams, see
1050 Documentation/networking/tcp-thin.rst
1054 tcp_limit_output_bytes - INTEGER
1055 Controls TCP Small Queue limit per tcp socket.
1056 TCP bulk sender tends to increase packets in flight until it
1057 gets losses notifications. With SNDBUF autotuning, this can
1058 result in a large amount of packets queued on the local machine
1059 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
1060 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
1061 limits the number of bytes on qdisc or device to reduce artificial
1062 RTT/cwnd and reduce bufferbloat.
1064 Default: 1048576 (16 * 65536)
1066 tcp_challenge_ack_limit - INTEGER
1067 Limits number of Challenge ACK sent per second, as recommended
1068 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
1069 Note that this per netns rate limit can allow some side channel
1070 attacks and probably should not be enabled.
1071 TCP stack implements per TCP socket limits anyway.
1072 Default: INT_MAX (unlimited)
1074 tcp_ehash_entries - INTEGER
1075 Show the number of hash buckets for TCP sockets in the current
1076 networking namespace.
1078 A negative value means the networking namespace does not own its
1079 hash buckets and shares the initial networking namespace's one.
1081 tcp_child_ehash_entries - INTEGER
1082 Control the number of hash buckets for TCP sockets in the child
1083 networking namespace, which must be set before clone() or unshare().
1085 If the value is not 0, the kernel uses a value rounded up to 2^n
1086 as the actual hash bucket size. 0 is a special value, meaning
1087 the child networking namespace will share the initial networking
1088 namespace's hash buckets.
1090 Note that the child will use the global one in case the kernel
1091 fails to allocate enough memory. In addition, the global hash
1092 buckets are spread over available NUMA nodes, but the allocation
1093 of the child hash table depends on the current process's NUMA
1094 policy, which could result in performance differences.
1096 Note also that the default value of tcp_max_tw_buckets and
1097 tcp_max_syn_backlog depend on the hash bucket size.
1099 Possible values: 0, 2^n (n: 0 - 24 (16Mi))
1103 tcp_plb_enabled - BOOLEAN
1104 If set and the underlying congestion control (e.g. DCTCP) supports
1105 and enables PLB feature, TCP PLB (Protective Load Balancing) is
1106 enabled. PLB is described in the following paper:
1107 https://doi.org/10.1145/3544216.3544226. Based on PLB parameters,
1108 upon sensing sustained congestion, TCP triggers a change in
1109 flow label field for outgoing IPv6 packets. A change in flow label
1110 field potentially changes the path of outgoing packets for switches
1111 that use ECMP/WCMP for routing.
1113 PLB changes socket txhash which results in a change in IPv6 Flow Label
1114 field, and currently no-op for IPv4 headers. It is possible
1115 to apply PLB for IPv4 with other network header fields (e.g. TCP
1116 or IPv4 options) or using encapsulation where outer header is used
1117 by switches to determine next hop. In either case, further host
1118 and switch side changes will be needed.
1120 When set, PLB assumes that congestion signal (e.g. ECN) is made
1121 available and used by congestion control module to estimate a
1122 congestion measure (e.g. ce_ratio). PLB needs a congestion measure to
1123 make repathing decisions.
1127 tcp_plb_idle_rehash_rounds - INTEGER
1128 Number of consecutive congested rounds (RTT) seen after which
1129 a rehash can be performed, given there are no packets in flight.
1130 This is referred to as M in PLB paper:
1131 https://doi.org/10.1145/3544216.3544226.
1133 Possible Values: 0 - 31
1137 tcp_plb_rehash_rounds - INTEGER
1138 Number of consecutive congested rounds (RTT) seen after which
1139 a forced rehash can be performed. Be careful when setting this
1140 parameter, as a small value increases the risk of retransmissions.
1141 This is referred to as N in PLB paper:
1142 https://doi.org/10.1145/3544216.3544226.
1144 Possible Values: 0 - 31
1148 tcp_plb_suspend_rto_sec - INTEGER
1149 Time, in seconds, to suspend PLB in event of an RTO. In order to avoid
1150 having PLB repath onto a connectivity "black hole", after an RTO a TCP
1151 connection suspends PLB repathing for a random duration between 1x and
1152 2x of this parameter. Randomness is added to avoid concurrent rehashing
1153 of multiple TCP connections. This should be set corresponding to the
1154 amount of time it takes to repair a failed link.
1156 Possible Values: 0 - 255
1160 tcp_plb_cong_thresh - INTEGER
1161 Fraction of packets marked with congestion over a round (RTT) to
1162 tag that round as congested. This is referred to as K in the PLB paper:
1163 https://doi.org/10.1145/3544216.3544226.
1165 The 0-1 fraction range is mapped to 0-256 range to avoid floating
1166 point operations. For example, 128 means that if at least 50% of
1167 the packets in a round were marked as congested then the round
1168 will be tagged as congested.
1170 Setting threshold to 0 means that PLB repaths every RTT regardless
1171 of congestion. This is not intended behavior for PLB and should be
1172 used only for experimentation purpose.
1174 Possible Values: 0 - 256
1181 udp_l3mdev_accept - BOOLEAN
1182 Enabling this option allows a "global" bound socket to work
1183 across L3 master domains (e.g., VRFs) with packets capable of
1184 being received regardless of the L3 domain in which they
1185 originated. Only valid when the kernel was compiled with
1186 CONFIG_NET_L3_MASTER_DEV.
1188 Default: 0 (disabled)
1190 udp_mem - vector of 3 INTEGERs: min, pressure, max
1191 Number of pages allowed for queueing by all UDP sockets.
1193 min: Number of pages allowed for queueing by all UDP sockets.
1195 pressure: This value was introduced to follow format of tcp_mem.
1197 max: This value was introduced to follow format of tcp_mem.
1199 Default is calculated at boot time from amount of available memory.
1201 udp_rmem_min - INTEGER
1202 Minimal size of receive buffer used by UDP sockets in moderation.
1203 Each UDP socket is able to use the size for receiving data, even if
1204 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1208 udp_wmem_min - INTEGER
1209 UDP does not have tx memory accounting and this tunable has no effect.
1211 udp_hash_entries - INTEGER
1212 Show the number of hash buckets for UDP sockets in the current
1213 networking namespace.
1215 A negative value means the networking namespace does not own its
1216 hash buckets and shares the initial networking namespace's one.
1218 udp_child_ehash_entries - INTEGER
1219 Control the number of hash buckets for UDP sockets in the child
1220 networking namespace, which must be set before clone() or unshare().
1222 If the value is not 0, the kernel uses a value rounded up to 2^n
1223 as the actual hash bucket size. 0 is a special value, meaning
1224 the child networking namespace will share the initial networking
1225 namespace's hash buckets.
1227 Note that the child will use the global one in case the kernel
1228 fails to allocate enough memory. In addition, the global hash
1229 buckets are spread over available NUMA nodes, but the allocation
1230 of the child hash table depends on the current process's NUMA
1231 policy, which could result in performance differences.
1233 Possible values: 0, 2^n (n: 7 (128) - 16 (64K))
1241 raw_l3mdev_accept - BOOLEAN
1242 Enabling this option allows a "global" bound socket to work
1243 across L3 master domains (e.g., VRFs) with packets capable of
1244 being received regardless of the L3 domain in which they
1245 originated. Only valid when the kernel was compiled with
1246 CONFIG_NET_L3_MASTER_DEV.
1248 Default: 1 (enabled)
1253 cipso_cache_enable - BOOLEAN
1254 If set, enable additions to and lookups from the CIPSO label mapping
1255 cache. If unset, additions are ignored and lookups always result in a
1256 miss. However, regardless of the setting the cache is still
1257 invalidated when required when means you can safely toggle this on and
1258 off and the cache will always be "safe".
1262 cipso_cache_bucket_size - INTEGER
1263 The CIPSO label cache consists of a fixed size hash table with each
1264 hash bucket containing a number of cache entries. This variable limits
1265 the number of entries in each hash bucket; the larger the value is, the
1266 more CIPSO label mappings that can be cached. When the number of
1267 entries in a given hash bucket reaches this limit adding new entries
1268 causes the oldest entry in the bucket to be removed to make room.
1272 cipso_rbm_optfmt - BOOLEAN
1273 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1274 the CIPSO draft specification (see Documentation/netlabel for details).
1275 This means that when set the CIPSO tag will be padded with empty
1276 categories in order to make the packet data 32-bit aligned.
1280 cipso_rbm_structvalid - BOOLEAN
1281 If set, do a very strict check of the CIPSO option when
1282 ip_options_compile() is called. If unset, relax the checks done during
1283 ip_options_compile(). Either way is "safe" as errors are caught else
1284 where in the CIPSO processing code but setting this to 0 (False) should
1285 result in less work (i.e. it should be faster) but could cause problems
1286 with other implementations that require strict checking.
1293 ip_local_port_range - 2 INTEGERS
1294 Defines the local port range that is used by TCP and UDP to
1295 choose the local port. The first number is the first, the
1296 second the last local port number.
1297 If possible, it is better these numbers have different parity
1298 (one even and one odd value).
1299 Must be greater than or equal to ip_unprivileged_port_start.
1300 The default values are 32768 and 60999 respectively.
1302 ip_local_reserved_ports - list of comma separated ranges
1303 Specify the ports which are reserved for known third-party
1304 applications. These ports will not be used by automatic port
1305 assignments (e.g. when calling connect() or bind() with port
1306 number 0). Explicit port allocation behavior is unchanged.
1308 The format used for both input and output is a comma separated
1309 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1310 10). Writing to the file will clear all previously reserved
1311 ports and update the current list with the one given in the
1314 Note that ip_local_port_range and ip_local_reserved_ports
1315 settings are independent and both are considered by the kernel
1316 when determining which ports are available for automatic port
1319 You can reserve ports which are not in the current
1320 ip_local_port_range, e.g.::
1322 $ cat /proc/sys/net/ipv4/ip_local_port_range
1324 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1327 although this is redundant. However such a setting is useful
1328 if later the port range is changed to a value that will
1329 include the reserved ports. Also keep in mind, that overlapping
1330 of these ranges may affect probability of selecting ephemeral
1331 ports which are right after block of reserved ports.
1335 ip_unprivileged_port_start - INTEGER
1336 This is a per-namespace sysctl. It defines the first
1337 unprivileged port in the network namespace. Privileged ports
1338 require root or CAP_NET_BIND_SERVICE in order to bind to them.
1339 To disable all privileged ports, set this to 0. They must not
1340 overlap with the ip_local_port_range.
1344 ip_nonlocal_bind - BOOLEAN
1345 If set, allows processes to bind() to non-local IP addresses,
1346 which can be quite useful - but may break some applications.
1350 ip_autobind_reuse - BOOLEAN
1351 By default, bind() does not select the ports automatically even if
1352 the new socket and all sockets bound to the port have SO_REUSEADDR.
1353 ip_autobind_reuse allows bind() to reuse the port and this is useful
1354 when you use bind()+connect(), but may break some applications.
1355 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1356 option should only be set by experts.
1359 ip_dynaddr - INTEGER
1360 If set non-zero, enables support for dynamic addresses.
1361 If set to a non-zero value larger than 1, a kernel log
1362 message will be printed when dynamic address rewriting
1367 ip_early_demux - BOOLEAN
1368 Optimize input packet processing down to one demux for
1369 certain kinds of local sockets. Currently we only do this
1370 for established TCP and connected UDP sockets.
1372 It may add an additional cost for pure routing workloads that
1373 reduces overall throughput, in such case you should disable it.
1377 ping_group_range - 2 INTEGERS
1378 Restrict ICMP_PROTO datagram sockets to users in the group range.
1379 The default is "1 0", meaning, that nobody (not even root) may
1380 create ping sockets. Setting it to "100 100" would grant permissions
1381 to the single group. "0 4294967294" would enable it for the world, "100
1382 4294967294" would enable it for the users, but not daemons.
1384 tcp_early_demux - BOOLEAN
1385 Enable early demux for established TCP sockets.
1389 udp_early_demux - BOOLEAN
1390 Enable early demux for connected UDP sockets. Disable this if
1391 your system could experience more unconnected load.
1395 icmp_echo_ignore_all - BOOLEAN
1396 If set non-zero, then the kernel will ignore all ICMP ECHO
1397 requests sent to it.
1401 icmp_echo_enable_probe - BOOLEAN
1402 If set to one, then the kernel will respond to RFC 8335 PROBE
1403 requests sent to it.
1407 icmp_echo_ignore_broadcasts - BOOLEAN
1408 If set non-zero, then the kernel will ignore all ICMP ECHO and
1409 TIMESTAMP requests sent to it via broadcast/multicast.
1413 icmp_ratelimit - INTEGER
1414 Limit the maximal rates for sending ICMP packets whose type matches
1415 icmp_ratemask (see below) to specific targets.
1416 0 to disable any limiting,
1417 otherwise the minimal space between responses in milliseconds.
1418 Note that another sysctl, icmp_msgs_per_sec limits the number
1419 of ICMP packets sent on all targets.
1423 icmp_msgs_per_sec - INTEGER
1424 Limit maximal number of ICMP packets sent per second from this host.
1425 Only messages whose type matches icmp_ratemask (see below) are
1426 controlled by this limit. For security reasons, the precise count
1427 of messages per second is randomized.
1431 icmp_msgs_burst - INTEGER
1432 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1433 while icmp_msgs_burst controls the burst size of these packets.
1434 For security reasons, the precise burst size is randomized.
1438 icmp_ratemask - INTEGER
1439 Mask made of ICMP types for which rates are being limited.
1441 Significant bits: IHGFEDCBA9876543210
1443 Default mask: 0000001100000011000 (6168)
1445 Bit definitions (see include/linux/icmp.h):
1447 = =========================
1449 3 Destination Unreachable [1]_
1450 4 Source Quench [1]_
1453 B Time Exceeded [1]_
1454 C Parameter Problem [1]_
1459 H Address Mask Request
1460 I Address Mask Reply
1461 = =========================
1463 .. [1] These are rate limited by default (see default mask above)
1465 icmp_ignore_bogus_error_responses - BOOLEAN
1466 Some routers violate RFC1122 by sending bogus responses to broadcast
1467 frames. Such violations are normally logged via a kernel warning.
1468 If this is set to TRUE, the kernel will not give such warnings, which
1469 will avoid log file clutter.
1473 icmp_errors_use_inbound_ifaddr - BOOLEAN
1475 If zero, icmp error messages are sent with the primary address of
1476 the exiting interface.
1478 If non-zero, the message will be sent with the primary address of
1479 the interface that received the packet that caused the icmp error.
1480 This is the behaviour many network administrators will expect from
1481 a router. And it can make debugging complicated network layouts
1484 Note that if no primary address exists for the interface selected,
1485 then the primary address of the first non-loopback interface that
1486 has one will be used regardless of this setting.
1490 igmp_max_memberships - INTEGER
1491 Change the maximum number of multicast groups we can subscribe to.
1494 Theoretical maximum value is bounded by having to send a membership
1495 report in a single datagram (i.e. the report can't span multiple
1496 datagrams, or risk confusing the switch and leaving groups you don't
1499 The number of supported groups 'M' is bounded by the number of group
1500 report entries you can fit into a single datagram of 65535 bytes.
1502 M = 65536-sizeof (ip header)/(sizeof(Group record))
1504 Group records are variable length, with a minimum of 12 bytes.
1505 So net.ipv4.igmp_max_memberships should not be set higher than:
1507 (65536-24) / 12 = 5459
1509 The value 5459 assumes no IP header options, so in practice
1510 this number may be lower.
1512 igmp_max_msf - INTEGER
1513 Maximum number of addresses allowed in the source filter list for a
1519 Controls the IGMP query robustness variable (see RFC2236 8.1).
1521 Default: 2 (as specified by RFC2236 8.1)
1523 Minimum: 1 (as specified by RFC6636 4.5)
1525 force_igmp_version - INTEGER
1526 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1527 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1528 Present timer expires.
1529 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1530 receive IGMPv2/v3 query.
1531 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1532 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1533 - 3 - Enforce to use IGMP version 3. The same react with default 0.
1537 this is not the same with force_mld_version because IGMPv3 RFC3376
1538 Security Considerations does not have clear description that we could
1539 ignore other version messages completely as MLDv2 RFC3810. So make
1540 this value as default 0 is recommended.
1542 ``conf/interface/*``
1543 changes special settings per interface (where
1544 interface" is the name of your network interface)
1547 is special, changes the settings for all interfaces
1549 log_martians - BOOLEAN
1550 Log packets with impossible addresses to kernel log.
1551 log_martians for the interface will be enabled if at least one of
1552 conf/{all,interface}/log_martians is set to TRUE,
1553 it will be disabled otherwise
1555 accept_redirects - BOOLEAN
1556 Accept ICMP redirect messages.
1557 accept_redirects for the interface will be enabled if:
1559 - both conf/{all,interface}/accept_redirects are TRUE in the case
1560 forwarding for the interface is enabled
1564 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1565 case forwarding for the interface is disabled
1567 accept_redirects for the interface will be disabled otherwise
1574 forwarding - BOOLEAN
1575 Enable IP forwarding on this interface. This controls whether packets
1576 received _on_ this interface can be forwarded.
1578 mc_forwarding - BOOLEAN
1579 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1580 and a multicast routing daemon is required.
1581 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1582 routing for the interface
1585 Integer value used to differentiate the devices by the medium they
1586 are attached to. Two devices can have different id values when
1587 the broadcast packets are received only on one of them.
1588 The default value 0 means that the device is the only interface
1589 to its medium, value of -1 means that medium is not known.
1591 Currently, it is used to change the proxy_arp behavior:
1592 the proxy_arp feature is enabled for packets forwarded between
1593 two devices attached to different media.
1598 proxy_arp for the interface will be enabled if at least one of
1599 conf/{all,interface}/proxy_arp is set to TRUE,
1600 it will be disabled otherwise
1602 proxy_arp_pvlan - BOOLEAN
1603 Private VLAN proxy arp.
1605 Basically allow proxy arp replies back to the same interface
1606 (from which the ARP request/solicitation was received).
1608 This is done to support (ethernet) switch features, like RFC
1609 3069, where the individual ports are NOT allowed to
1610 communicate with each other, but they are allowed to talk to
1611 the upstream router. As described in RFC 3069, it is possible
1612 to allow these hosts to communicate through the upstream
1613 router by proxy_arp'ing. Don't need to be used together with
1616 This technology is known by different names:
1618 In RFC 3069 it is called VLAN Aggregation.
1619 Cisco and Allied Telesyn call it Private VLAN.
1620 Hewlett-Packard call it Source-Port filtering or port-isolation.
1621 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1623 proxy_delay - INTEGER
1624 Delay proxy response.
1626 Delay response to a neighbor solicitation when proxy_arp
1627 or proxy_ndp is enabled. A random value between [0, proxy_delay)
1628 will be chosen, setting to zero means reply with no delay.
1629 Value in jiffies. Defaults to 80.
1631 shared_media - BOOLEAN
1632 Send(router) or accept(host) RFC1620 shared media redirects.
1633 Overrides secure_redirects.
1635 shared_media for the interface will be enabled if at least one of
1636 conf/{all,interface}/shared_media is set to TRUE,
1637 it will be disabled otherwise
1641 secure_redirects - BOOLEAN
1642 Accept ICMP redirect messages only to gateways listed in the
1643 interface's current gateway list. Even if disabled, RFC1122 redirect
1646 Overridden by shared_media.
1648 secure_redirects for the interface will be enabled if at least one of
1649 conf/{all,interface}/secure_redirects is set to TRUE,
1650 it will be disabled otherwise
1654 send_redirects - BOOLEAN
1655 Send redirects, if router.
1657 send_redirects for the interface will be enabled if at least one of
1658 conf/{all,interface}/send_redirects is set to TRUE,
1659 it will be disabled otherwise
1663 bootp_relay - BOOLEAN
1664 Accept packets with source address 0.b.c.d destined
1665 not to this host as local ones. It is supposed, that
1666 BOOTP relay daemon will catch and forward such packets.
1667 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1672 Not Implemented Yet.
1674 accept_source_route - BOOLEAN
1675 Accept packets with SRR option.
1676 conf/all/accept_source_route must also be set to TRUE to accept packets
1677 with SRR option on the interface
1684 accept_local - BOOLEAN
1685 Accept packets with local source addresses. In combination with
1686 suitable routing, this can be used to direct packets between two
1687 local interfaces over the wire and have them accepted properly.
1690 route_localnet - BOOLEAN
1691 Do not consider loopback addresses as martian source or destination
1692 while routing. This enables the use of 127/8 for local routing purposes.
1697 - 0 - No source validation.
1698 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1699 Each incoming packet is tested against the FIB and if the interface
1700 is not the best reverse path the packet check will fail.
1701 By default failed packets are discarded.
1702 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1703 Each incoming packet's source address is also tested against the FIB
1704 and if the source address is not reachable via any interface
1705 the packet check will fail.
1707 Current recommended practice in RFC3704 is to enable strict mode
1708 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1709 or other complicated routing, then loose mode is recommended.
1711 The max value from conf/{all,interface}/rp_filter is used
1712 when doing source validation on the {interface}.
1714 Default value is 0. Note that some distributions enable it
1717 src_valid_mark - BOOLEAN
1718 - 0 - The fwmark of the packet is not included in reverse path
1719 route lookup. This allows for asymmetric routing configurations
1720 utilizing the fwmark in only one direction, e.g., transparent
1723 - 1 - The fwmark of the packet is included in reverse path route
1724 lookup. This permits rp_filter to function when the fwmark is
1725 used for routing traffic in both directions.
1727 This setting also affects the utilization of fmwark when
1728 performing source address selection for ICMP replies, or
1729 determining addresses stored for the IPOPT_TS_TSANDADDR and
1730 IPOPT_RR IP options.
1732 The max value from conf/{all,interface}/src_valid_mark is used.
1736 arp_filter - BOOLEAN
1737 - 1 - Allows you to have multiple network interfaces on the same
1738 subnet, and have the ARPs for each interface be answered
1739 based on whether or not the kernel would route a packet from
1740 the ARP'd IP out that interface (therefore you must use source
1741 based routing for this to work). In other words it allows control
1742 of which cards (usually 1) will respond to an arp request.
1744 - 0 - (default) The kernel can respond to arp requests with addresses
1745 from other interfaces. This may seem wrong but it usually makes
1746 sense, because it increases the chance of successful communication.
1747 IP addresses are owned by the complete host on Linux, not by
1748 particular interfaces. Only for more complex setups like load-
1749 balancing, does this behaviour cause problems.
1751 arp_filter for the interface will be enabled if at least one of
1752 conf/{all,interface}/arp_filter is set to TRUE,
1753 it will be disabled otherwise
1755 arp_announce - INTEGER
1756 Define different restriction levels for announcing the local
1757 source IP address from IP packets in ARP requests sent on
1760 - 0 - (default) Use any local address, configured on any interface
1761 - 1 - Try to avoid local addresses that are not in the target's
1762 subnet for this interface. This mode is useful when target
1763 hosts reachable via this interface require the source IP
1764 address in ARP requests to be part of their logical network
1765 configured on the receiving interface. When we generate the
1766 request we will check all our subnets that include the
1767 target IP and will preserve the source address if it is from
1768 such subnet. If there is no such subnet we select source
1769 address according to the rules for level 2.
1770 - 2 - Always use the best local address for this target.
1771 In this mode we ignore the source address in the IP packet
1772 and try to select local address that we prefer for talks with
1773 the target host. Such local address is selected by looking
1774 for primary IP addresses on all our subnets on the outgoing
1775 interface that include the target IP address. If no suitable
1776 local address is found we select the first local address
1777 we have on the outgoing interface or on all other interfaces,
1778 with the hope we will receive reply for our request and
1779 even sometimes no matter the source IP address we announce.
1781 The max value from conf/{all,interface}/arp_announce is used.
1783 Increasing the restriction level gives more chance for
1784 receiving answer from the resolved target while decreasing
1785 the level announces more valid sender's information.
1787 arp_ignore - INTEGER
1788 Define different modes for sending replies in response to
1789 received ARP requests that resolve local target IP addresses:
1791 - 0 - (default): reply for any local target IP address, configured
1793 - 1 - reply only if the target IP address is local address
1794 configured on the incoming interface
1795 - 2 - reply only if the target IP address is local address
1796 configured on the incoming interface and both with the
1797 sender's IP address are part from same subnet on this interface
1798 - 3 - do not reply for local addresses configured with scope host,
1799 only resolutions for global and link addresses are replied
1801 - 8 - do not reply for all local addresses
1803 The max value from conf/{all,interface}/arp_ignore is used
1804 when ARP request is received on the {interface}
1806 arp_notify - BOOLEAN
1807 Define mode for notification of address and device changes.
1809 == ==========================================================
1810 0 (default): do nothing
1811 1 Generate gratuitous arp requests when device is brought up
1812 or hardware address changes.
1813 == ==========================================================
1815 arp_accept - INTEGER
1816 Define behavior for accepting gratuitous ARP (garp) frames from devices
1817 that are not already present in the ARP table:
1819 - 0 - don't create new entries in the ARP table
1820 - 1 - create new entries in the ARP table
1821 - 2 - create new entries only if the source IP address is in the same
1822 subnet as an address configured on the interface that received the
1825 Both replies and requests type gratuitous arp will trigger the
1826 ARP table to be updated, if this setting is on.
1828 If the ARP table already contains the IP address of the
1829 gratuitous arp frame, the arp table will be updated regardless
1830 if this setting is on or off.
1832 arp_evict_nocarrier - BOOLEAN
1833 Clears the ARP cache on NOCARRIER events. This option is important for
1834 wireless devices where the ARP cache should not be cleared when roaming
1835 between access points on the same network. In most cases this should
1836 remain as the default (1).
1838 - 1 - (default): Clear the ARP cache on NOCARRIER events
1839 - 0 - Do not clear ARP cache on NOCARRIER events
1841 mcast_solicit - INTEGER
1842 The maximum number of multicast probes in INCOMPLETE state,
1843 when the associated hardware address is unknown. Defaults
1846 ucast_solicit - INTEGER
1847 The maximum number of unicast probes in PROBE state, when
1848 the hardware address is being reconfirmed. Defaults to 3.
1850 app_solicit - INTEGER
1851 The maximum number of probes to send to the user space ARP daemon
1852 via netlink before dropping back to multicast probes (see
1853 mcast_resolicit). Defaults to 0.
1855 mcast_resolicit - INTEGER
1856 The maximum number of multicast probes after unicast and
1857 app probes in PROBE state. Defaults to 0.
1859 disable_policy - BOOLEAN
1860 Disable IPSEC policy (SPD) for this interface
1862 disable_xfrm - BOOLEAN
1863 Disable IPSEC encryption on this interface, whatever the policy
1865 igmpv2_unsolicited_report_interval - INTEGER
1866 The interval in milliseconds in which the next unsolicited
1867 IGMPv1 or IGMPv2 report retransmit will take place.
1869 Default: 10000 (10 seconds)
1871 igmpv3_unsolicited_report_interval - INTEGER
1872 The interval in milliseconds in which the next unsolicited
1873 IGMPv3 report retransmit will take place.
1875 Default: 1000 (1 seconds)
1877 ignore_routes_with_linkdown - BOOLEAN
1878 Ignore routes whose link is down when performing a FIB lookup.
1880 promote_secondaries - BOOLEAN
1881 When a primary IP address is removed from this interface
1882 promote a corresponding secondary IP address instead of
1883 removing all the corresponding secondary IP addresses.
1885 drop_unicast_in_l2_multicast - BOOLEAN
1886 Drop any unicast IP packets that are received in link-layer
1887 multicast (or broadcast) frames.
1889 This behavior (for multicast) is actually a SHOULD in RFC
1890 1122, but is disabled by default for compatibility reasons.
1894 drop_gratuitous_arp - BOOLEAN
1895 Drop all gratuitous ARP frames, for example if there's a known
1896 good ARP proxy on the network and such frames need not be used
1897 (or in the case of 802.11, must not be used to prevent attacks.)
1903 Allows you to write a number, which can be used as required.
1907 xfrm4_gc_thresh - INTEGER
1908 (Obsolete since linux-4.14)
1909 The threshold at which we will start garbage collecting for IPv4
1910 destination cache entries. At twice this value the system will
1911 refuse new allocations.
1913 igmp_link_local_mcast_reports - BOOLEAN
1914 Enable IGMP reports for link local multicast groups in the
1920 kuznet@ms2.inr.ac.ru
1927 delon.nicolas@wanadoo.fr
1932 /proc/sys/net/ipv6/* Variables
1933 ==============================
1935 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1936 apply to IPv6 [XXX?].
1938 bindv6only - BOOLEAN
1939 Default value for IPV6_V6ONLY socket option,
1940 which restricts use of the IPv6 socket to IPv6 communication
1943 - TRUE: disable IPv4-mapped address feature
1944 - FALSE: enable IPv4-mapped address feature
1946 Default: FALSE (as specified in RFC3493)
1948 flowlabel_consistency - BOOLEAN
1949 Protect the consistency (and unicity) of flow label.
1950 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1958 auto_flowlabels - INTEGER
1959 Automatically generate flow labels based on a flow hash of the
1960 packet. This allows intermediate devices, such as routers, to
1961 identify packet flows for mechanisms like Equal Cost Multipath
1962 Routing (see RFC 6438).
1964 = ===========================================================
1965 0 automatic flow labels are completely disabled
1966 1 automatic flow labels are enabled by default, they can be
1967 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1969 2 automatic flow labels are allowed, they may be enabled on a
1970 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1971 3 automatic flow labels are enabled and enforced, they cannot
1972 be disabled by the socket option
1973 = ===========================================================
1977 flowlabel_state_ranges - BOOLEAN
1978 Split the flow label number space into two ranges. 0-0x7FFFF is
1979 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1980 is reserved for stateless flow labels as described in RFC6437.
1987 flowlabel_reflect - INTEGER
1988 Control flow label reflection. Needed for Path MTU
1989 Discovery to work with Equal Cost Multipath Routing in anycast
1990 environments. See RFC 7690 and:
1991 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1995 - 1: enabled for established flows
1997 Note that this prevents automatic flowlabel changes, as done
1998 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
1999 and "tcp: Change txhash on every SYN and RTO retransmit"
2001 - 2: enabled for TCP RESET packets (no active listener)
2002 If set, a RST packet sent in response to a SYN packet on a closed
2003 port will reflect the incoming flow label.
2005 - 4: enabled for ICMPv6 echo reply messages.
2009 fib_multipath_hash_policy - INTEGER
2010 Controls which hash policy to use for multipath routes.
2012 Default: 0 (Layer 3)
2016 - 0 - Layer 3 (source and destination addresses plus flow label)
2017 - 1 - Layer 4 (standard 5-tuple)
2018 - 2 - Layer 3 or inner Layer 3 if present
2019 - 3 - Custom multipath hash. Fields used for multipath hash calculation
2020 are determined by fib_multipath_hash_fields sysctl
2022 fib_multipath_hash_fields - UNSIGNED INTEGER
2023 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
2024 fields used for multipath hash calculation are determined by this
2027 This value is a bitmask which enables various fields for multipath hash
2030 Possible fields are:
2032 ====== ============================
2033 0x0001 Source IP address
2034 0x0002 Destination IP address
2038 0x0020 Destination port
2039 0x0040 Inner source IP address
2040 0x0080 Inner destination IP address
2041 0x0100 Inner IP protocol
2042 0x0200 Inner Flow Label
2043 0x0400 Inner source port
2044 0x0800 Inner destination port
2045 ====== ============================
2047 Default: 0x0007 (source IP, destination IP and IP protocol)
2049 anycast_src_echo_reply - BOOLEAN
2050 Controls the use of anycast addresses as source addresses for ICMPv6
2058 idgen_delay - INTEGER
2059 Controls the delay in seconds after which time to retry
2060 privacy stable address generation if a DAD conflict is
2063 Default: 1 (as specified in RFC7217)
2065 idgen_retries - INTEGER
2066 Controls the number of retries to generate a stable privacy
2067 address if a DAD conflict is detected.
2069 Default: 3 (as specified in RFC7217)
2072 Controls the MLD query robustness variable (see RFC3810 9.1).
2074 Default: 2 (as specified by RFC3810 9.1)
2076 Minimum: 1 (as specified by RFC6636 4.5)
2078 max_dst_opts_number - INTEGER
2079 Maximum number of non-padding TLVs allowed in a Destination
2080 options extension header. If this value is less than zero
2081 then unknown options are disallowed and the number of known
2082 TLVs allowed is the absolute value of this number.
2086 max_hbh_opts_number - INTEGER
2087 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
2088 options extension header. If this value is less than zero
2089 then unknown options are disallowed and the number of known
2090 TLVs allowed is the absolute value of this number.
2094 max_dst_opts_length - INTEGER
2095 Maximum length allowed for a Destination options extension
2098 Default: INT_MAX (unlimited)
2100 max_hbh_length - INTEGER
2101 Maximum length allowed for a Hop-by-Hop options extension
2104 Default: INT_MAX (unlimited)
2106 skip_notify_on_dev_down - BOOLEAN
2107 Controls whether an RTM_DELROUTE message is generated for routes
2108 removed when a device is taken down or deleted. IPv4 does not
2109 generate this message; IPv6 does by default. Setting this sysctl
2110 to true skips the message, making IPv4 and IPv6 on par in relying
2111 on userspace caches to track link events and evict routes.
2113 Default: false (generate message)
2115 nexthop_compat_mode - BOOLEAN
2116 New nexthop API provides a means for managing nexthops independent of
2117 prefixes. Backwards compatibility with old route format is enabled by
2118 default which means route dumps and notifications contain the new
2119 nexthop attribute but also the full, expanded nexthop definition.
2120 Further, updates or deletes of a nexthop configuration generate route
2121 notifications for each fib entry using the nexthop. Once a system
2122 understands the new API, this sysctl can be disabled to achieve full
2123 performance benefits of the new API by disabling the nexthop expansion
2124 and extraneous notifications.
2125 Default: true (backward compat mode)
2127 fib_notify_on_flag_change - INTEGER
2128 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
2129 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
2131 After installing a route to the kernel, user space receives an
2132 acknowledgment, which means the route was installed in the kernel,
2133 but not necessarily in hardware.
2134 It is also possible for a route already installed in hardware to change
2135 its action and therefore its flags. For example, a host route that is
2136 trapping packets can be "promoted" to perform decapsulation following
2137 the installation of an IPinIP/VXLAN tunnel.
2138 The notifications will indicate to user-space the state of the route.
2140 Default: 0 (Do not emit notifications.)
2144 - 0 - Do not emit notifications.
2145 - 1 - Emit notifications.
2146 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
2149 Define the IOAM id of this node. Uses only 24 bits out of 32 in total.
2156 ioam6_id_wide - LONG INTEGER
2157 Define the wide IOAM id of this node. Uses only 56 bits out of 64 in
2158 total. Can be different from ioam6_id.
2161 Max: 0xFFFFFFFFFFFFFF
2163 Default: 0xFFFFFFFFFFFFFF
2167 ip6frag_high_thresh - INTEGER
2168 Maximum memory used to reassemble IPv6 fragments. When
2169 ip6frag_high_thresh bytes of memory is allocated for this purpose,
2170 the fragment handler will toss packets until ip6frag_low_thresh
2173 ip6frag_low_thresh - INTEGER
2174 See ip6frag_high_thresh
2176 ip6frag_time - INTEGER
2177 Time in seconds to keep an IPv6 fragment in memory.
2180 Change the interface-specific default settings.
2182 These settings would be used during creating new interfaces.
2186 Change all the interface-specific settings.
2188 [XXX: Other special features than forwarding?]
2190 conf/all/disable_ipv6 - BOOLEAN
2191 Changing this value is same as changing ``conf/default/disable_ipv6``
2192 setting and also all per-interface ``disable_ipv6`` settings to the same
2195 Reading this value does not have any particular meaning. It does not say
2196 whether IPv6 support is enabled or disabled. Returned value can be 1
2197 also in the case when some interface has ``disable_ipv6`` set to 0 and
2198 has configured IPv6 addresses.
2200 conf/all/forwarding - BOOLEAN
2201 Enable global IPv6 forwarding between all interfaces.
2203 IPv4 and IPv6 work differently here; e.g. netfilter must be used
2204 to control which interfaces may forward packets and which not.
2206 This also sets all interfaces' Host/Router setting
2207 'forwarding' to the specified value. See below for details.
2209 This referred to as global forwarding.
2214 fwmark_reflect - BOOLEAN
2215 Controls the fwmark of kernel-generated IPv6 reply packets that are not
2216 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
2217 If unset, these packets have a fwmark of zero. If set, they have the
2218 fwmark of the packet they are replying to.
2222 ``conf/interface/*``:
2223 Change special settings per interface.
2225 The functional behaviour for certain settings is different
2226 depending on whether local forwarding is enabled or not.
2229 Accept Router Advertisements; autoconfigure using them.
2231 It also determines whether or not to transmit Router
2232 Solicitations. If and only if the functional setting is to
2233 accept Router Advertisements, Router Solicitations will be
2236 Possible values are:
2238 == ===========================================================
2239 0 Do not accept Router Advertisements.
2240 1 Accept Router Advertisements if forwarding is disabled.
2241 2 Overrule forwarding behaviour. Accept Router Advertisements
2242 even if forwarding is enabled.
2243 == ===========================================================
2247 - enabled if local forwarding is disabled.
2248 - disabled if local forwarding is enabled.
2250 accept_ra_defrtr - BOOLEAN
2251 Learn default router in Router Advertisement.
2255 - enabled if accept_ra is enabled.
2256 - disabled if accept_ra is disabled.
2258 ra_defrtr_metric - UNSIGNED INTEGER
2259 Route metric for default route learned in Router Advertisement. This value
2260 will be assigned as metric for the default route learned via IPv6 Router
2261 Advertisement. Takes affect only if accept_ra_defrtr is enabled.
2266 Default: IP6_RT_PRIO_USER i.e. 1024.
2268 accept_ra_from_local - BOOLEAN
2269 Accept RA with source-address that is found on local machine
2270 if the RA is otherwise proper and able to be accepted.
2272 Default is to NOT accept these as it may be an un-intended
2277 - enabled if accept_ra_from_local is enabled
2278 on a specific interface.
2279 - disabled if accept_ra_from_local is disabled
2280 on a specific interface.
2282 accept_ra_min_hop_limit - INTEGER
2283 Minimum hop limit Information in Router Advertisement.
2285 Hop limit Information in Router Advertisement less than this
2286 variable shall be ignored.
2290 accept_ra_pinfo - BOOLEAN
2291 Learn Prefix Information in Router Advertisement.
2295 - enabled if accept_ra is enabled.
2296 - disabled if accept_ra is disabled.
2298 accept_ra_rt_info_min_plen - INTEGER
2299 Minimum prefix length of Route Information in RA.
2301 Route Information w/ prefix smaller than this variable shall
2306 * 0 if accept_ra_rtr_pref is enabled.
2307 * -1 if accept_ra_rtr_pref is disabled.
2309 accept_ra_rt_info_max_plen - INTEGER
2310 Maximum prefix length of Route Information in RA.
2312 Route Information w/ prefix larger than this variable shall
2317 * 0 if accept_ra_rtr_pref is enabled.
2318 * -1 if accept_ra_rtr_pref is disabled.
2320 accept_ra_rtr_pref - BOOLEAN
2321 Accept Router Preference in RA.
2325 - enabled if accept_ra is enabled.
2326 - disabled if accept_ra is disabled.
2328 accept_ra_mtu - BOOLEAN
2329 Apply the MTU value specified in RA option 5 (RFC4861). If
2330 disabled, the MTU specified in the RA will be ignored.
2334 - enabled if accept_ra is enabled.
2335 - disabled if accept_ra is disabled.
2337 accept_redirects - BOOLEAN
2342 - enabled if local forwarding is disabled.
2343 - disabled if local forwarding is enabled.
2345 accept_source_route - INTEGER
2346 Accept source routing (routing extension header).
2348 - >= 0: Accept only routing header type 2.
2349 - < 0: Do not accept routing header.
2354 Autoconfigure addresses using Prefix Information in Router
2359 - enabled if accept_ra_pinfo is enabled.
2360 - disabled if accept_ra_pinfo is disabled.
2362 dad_transmits - INTEGER
2363 The amount of Duplicate Address Detection probes to send.
2367 forwarding - INTEGER
2368 Configure interface-specific Host/Router behaviour.
2372 It is recommended to have the same setting on all
2373 interfaces; mixed router/host scenarios are rather uncommon.
2375 Possible values are:
2377 - 0 Forwarding disabled
2378 - 1 Forwarding enabled
2382 By default, Host behaviour is assumed. This means:
2384 1. IsRouter flag is not set in Neighbour Advertisements.
2385 2. If accept_ra is TRUE (default), transmit Router
2387 3. If accept_ra is TRUE (default), accept Router
2388 Advertisements (and do autoconfiguration).
2389 4. If accept_redirects is TRUE (default), accept Redirects.
2393 If local forwarding is enabled, Router behaviour is assumed.
2394 This means exactly the reverse from the above:
2396 1. IsRouter flag is set in Neighbour Advertisements.
2397 2. Router Solicitations are not sent unless accept_ra is 2.
2398 3. Router Advertisements are ignored unless accept_ra is 2.
2399 4. Redirects are ignored.
2401 Default: 0 (disabled) if global forwarding is disabled (default),
2402 otherwise 1 (enabled).
2405 Default Hop Limit to set.
2410 Default Maximum Transfer Unit
2412 Default: 1280 (IPv6 required minimum)
2414 ip_nonlocal_bind - BOOLEAN
2415 If set, allows processes to bind() to non-local IPv6 addresses,
2416 which can be quite useful - but may break some applications.
2420 router_probe_interval - INTEGER
2421 Minimum interval (in seconds) between Router Probing described
2426 router_solicitation_delay - INTEGER
2427 Number of seconds to wait after interface is brought up
2428 before sending Router Solicitations.
2432 router_solicitation_interval - INTEGER
2433 Number of seconds to wait between Router Solicitations.
2437 router_solicitations - INTEGER
2438 Number of Router Solicitations to send until assuming no
2439 routers are present.
2443 use_oif_addrs_only - BOOLEAN
2444 When enabled, the candidate source addresses for destinations
2445 routed via this interface are restricted to the set of addresses
2446 configured on this interface (vis. RFC 6724, section 4).
2450 use_tempaddr - INTEGER
2451 Preference for Privacy Extensions (RFC3041).
2453 * <= 0 : disable Privacy Extensions
2454 * == 1 : enable Privacy Extensions, but prefer public
2455 addresses over temporary addresses.
2456 * > 1 : enable Privacy Extensions and prefer temporary
2457 addresses over public addresses.
2461 * 0 (for most devices)
2462 * -1 (for point-to-point devices and loopback devices)
2464 temp_valid_lft - INTEGER
2465 valid lifetime (in seconds) for temporary addresses.
2467 Default: 172800 (2 days)
2469 temp_prefered_lft - INTEGER
2470 Preferred lifetime (in seconds) for temporary addresses.
2472 Default: 86400 (1 day)
2474 keep_addr_on_down - INTEGER
2475 Keep all IPv6 addresses on an interface down event. If set static
2476 global addresses with no expiration time are not flushed.
2479 * 0 : system default
2482 Default: 0 (addresses are removed)
2484 max_desync_factor - INTEGER
2485 Maximum value for DESYNC_FACTOR, which is a random value
2486 that ensures that clients don't synchronize with each
2487 other and generate new addresses at exactly the same time.
2488 value is in seconds.
2492 regen_max_retry - INTEGER
2493 Number of attempts before give up attempting to generate
2494 valid temporary addresses.
2498 max_addresses - INTEGER
2499 Maximum number of autoconfigured addresses per interface. Setting
2500 to zero disables the limitation. It is not recommended to set this
2501 value too large (or to zero) because it would be an easy way to
2502 crash the kernel by allowing too many addresses to be created.
2506 disable_ipv6 - BOOLEAN
2507 Disable IPv6 operation. If accept_dad is set to 2, this value
2508 will be dynamically set to TRUE if DAD fails for the link-local
2511 Default: FALSE (enable IPv6 operation)
2513 When this value is changed from 1 to 0 (IPv6 is being enabled),
2514 it will dynamically create a link-local address on the given
2515 interface and start Duplicate Address Detection, if necessary.
2517 When this value is changed from 0 to 1 (IPv6 is being disabled),
2518 it will dynamically delete all addresses and routes on the given
2519 interface. From now on it will not possible to add addresses/routes
2520 to the selected interface.
2522 accept_dad - INTEGER
2523 Whether to accept DAD (Duplicate Address Detection).
2525 == ==============================================================
2527 1 Enable DAD (default)
2528 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate
2529 link-local address has been found.
2530 == ==============================================================
2532 DAD operation and mode on a given interface will be selected according
2533 to the maximum value of conf/{all,interface}/accept_dad.
2535 force_tllao - BOOLEAN
2536 Enable sending the target link-layer address option even when
2537 responding to a unicast neighbor solicitation.
2541 Quoting from RFC 2461, section 4.4, Target link-layer address:
2543 "The option MUST be included for multicast solicitations in order to
2544 avoid infinite Neighbor Solicitation "recursion" when the peer node
2545 does not have a cache entry to return a Neighbor Advertisements
2546 message. When responding to unicast solicitations, the option can be
2547 omitted since the sender of the solicitation has the correct link-
2548 layer address; otherwise it would not have be able to send the unicast
2549 solicitation in the first place. However, including the link-layer
2550 address in this case adds little overhead and eliminates a potential
2551 race condition where the sender deletes the cached link-layer address
2552 prior to receiving a response to a previous solicitation."
2554 ndisc_notify - BOOLEAN
2555 Define mode for notification of address and device changes.
2557 * 0 - (default): do nothing
2558 * 1 - Generate unsolicited neighbour advertisements when device is brought
2559 up or hardware address changes.
2561 ndisc_tclass - INTEGER
2562 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
2563 Discovery (Router Solicitation, Router Advertisement, Neighbor
2564 Solicitation, Neighbor Advertisement, Redirect) messages.
2565 These 8 bits can be interpreted as 6 high order bits holding the DSCP
2566 value and 2 low order bits representing ECN (which you probably want
2571 ndisc_evict_nocarrier - BOOLEAN
2572 Clears the neighbor discovery table on NOCARRIER events. This option is
2573 important for wireless devices where the neighbor discovery cache should
2574 not be cleared when roaming between access points on the same network.
2575 In most cases this should remain as the default (1).
2577 - 1 - (default): Clear neighbor discover cache on NOCARRIER events.
2578 - 0 - Do not clear neighbor discovery cache on NOCARRIER events.
2580 mldv1_unsolicited_report_interval - INTEGER
2581 The interval in milliseconds in which the next unsolicited
2582 MLDv1 report retransmit will take place.
2584 Default: 10000 (10 seconds)
2586 mldv2_unsolicited_report_interval - INTEGER
2587 The interval in milliseconds in which the next unsolicited
2588 MLDv2 report retransmit will take place.
2590 Default: 1000 (1 second)
2592 force_mld_version - INTEGER
2593 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
2594 * 1 - Enforce to use MLD version 1
2595 * 2 - Enforce to use MLD version 2
2597 suppress_frag_ndisc - INTEGER
2598 Control RFC 6980 (Security Implications of IPv6 Fragmentation
2599 with IPv6 Neighbor Discovery) behavior:
2601 * 1 - (default) discard fragmented neighbor discovery packets
2602 * 0 - allow fragmented neighbor discovery packets
2604 optimistic_dad - BOOLEAN
2605 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
2607 * 0: disabled (default)
2610 Optimistic Duplicate Address Detection for the interface will be enabled
2611 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
2612 it will be disabled otherwise.
2614 use_optimistic - BOOLEAN
2615 If enabled, do not classify optimistic addresses as deprecated during
2616 source address selection. Preferred addresses will still be chosen
2617 before optimistic addresses, subject to other ranking in the source
2618 address selection algorithm.
2620 * 0: disabled (default)
2623 This will be enabled if at least one of
2624 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
2626 stable_secret - IPv6 address
2627 This IPv6 address will be used as a secret to generate IPv6
2628 addresses for link-local addresses and autoconfigured
2629 ones. All addresses generated after setting this secret will
2630 be stable privacy ones by default. This can be changed via the
2631 addrgenmode ip-link. conf/default/stable_secret is used as the
2632 secret for the namespace, the interface specific ones can
2633 overwrite that. Writes to conf/all/stable_secret are refused.
2635 It is recommended to generate this secret during installation
2636 of a system and keep it stable after that.
2638 By default the stable secret is unset.
2640 addr_gen_mode - INTEGER
2641 Defines how link-local and autoconf addresses are generated.
2643 = =================================================================
2644 0 generate address based on EUI64 (default)
2645 1 do no generate a link-local address, use EUI64 for addresses
2646 generated from autoconf
2647 2 generate stable privacy addresses, using the secret from
2648 stable_secret (RFC7217)
2649 3 generate stable privacy addresses, using a random secret if unset
2650 = =================================================================
2652 drop_unicast_in_l2_multicast - BOOLEAN
2653 Drop any unicast IPv6 packets that are received in link-layer
2654 multicast (or broadcast) frames.
2656 By default this is turned off.
2658 drop_unsolicited_na - BOOLEAN
2659 Drop all unsolicited neighbor advertisements, for example if there's
2660 a known good NA proxy on the network and such frames need not be used
2661 (or in the case of 802.11, must not be used to prevent attacks.)
2663 By default this is turned off.
2665 accept_untracked_na - INTEGER
2666 Define behavior for accepting neighbor advertisements from devices that
2667 are absent in the neighbor cache:
2669 - 0 - (default) Do not accept unsolicited and untracked neighbor
2672 - 1 - Add a new neighbor cache entry in STALE state for routers on
2673 receiving a neighbor advertisement (either solicited or unsolicited)
2674 with target link-layer address option specified if no neighbor entry
2675 is already present for the advertised IPv6 address. Without this knob,
2676 NAs received for untracked addresses (absent in neighbor cache) are
2679 This is as per router-side behavior documented in RFC9131.
2681 This has lower precedence than drop_unsolicited_na.
2683 This will optimize the return path for the initial off-link
2684 communication that is initiated by a directly connected host, by
2685 ensuring that the first-hop router which turns on this setting doesn't
2686 have to buffer the initial return packets to do neighbor-solicitation.
2687 The prerequisite is that the host is configured to send unsolicited
2688 neighbor advertisements on interface bringup. This setting should be
2689 used in conjunction with the ndisc_notify setting on the host to
2690 satisfy this prerequisite.
2692 - 2 - Extend option (1) to add a new neighbor cache entry only if the
2693 source IP address is in the same subnet as an address configured on
2694 the interface that received the neighbor advertisement.
2696 enhanced_dad - BOOLEAN
2697 Include a nonce option in the IPv6 neighbor solicitation messages used for
2698 duplicate address detection per RFC7527. A received DAD NS will only signal
2699 a duplicate address if the nonce is different. This avoids any false
2700 detection of duplicates due to loopback of the NS messages that we send.
2701 The nonce option will be sent on an interface unless both of
2702 conf/{all,interface}/enhanced_dad are set to FALSE.
2710 Limit the maximal rates for sending ICMPv6 messages.
2712 0 to disable any limiting,
2713 otherwise the minimal space between responses in milliseconds.
2717 ratemask - list of comma separated ranges
2718 For ICMPv6 message types matching the ranges in the ratemask, limit
2719 the sending of the message according to ratelimit parameter.
2721 The format used for both input and output is a comma separated
2722 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
2723 129). Writing to the file will clear all previous ranges of ICMPv6
2724 message types and update the current list with the input.
2726 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
2727 for numerical values of ICMPv6 message types, e.g. echo request is 128
2728 and echo reply is 129.
2730 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
2732 echo_ignore_all - BOOLEAN
2733 If set non-zero, then the kernel will ignore all ICMP ECHO
2734 requests sent to it over the IPv6 protocol.
2738 echo_ignore_multicast - BOOLEAN
2739 If set non-zero, then the kernel will ignore all ICMP ECHO
2740 requests sent to it over the IPv6 protocol via multicast.
2744 echo_ignore_anycast - BOOLEAN
2745 If set non-zero, then the kernel will ignore all ICMP ECHO
2746 requests sent to it over the IPv6 protocol destined to anycast address.
2750 error_anycast_as_unicast - BOOLEAN
2751 If set to 1, then the kernel will respond with ICMP Errors
2752 resulting from requests sent to it over the IPv6 protocol destined
2753 to anycast address essentially treating anycast as unicast.
2757 xfrm6_gc_thresh - INTEGER
2758 (Obsolete since linux-4.14)
2759 The threshold at which we will start garbage collecting for IPv6
2760 destination cache entries. At twice this value the system will
2761 refuse new allocations.
2765 Pekka Savola <pekkas@netcore.fi>
2766 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2769 /proc/sys/net/bridge/* Variables:
2770 =================================
2772 bridge-nf-call-arptables - BOOLEAN
2773 - 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2778 bridge-nf-call-iptables - BOOLEAN
2779 - 1 : pass bridged IPv4 traffic to iptables' chains.
2784 bridge-nf-call-ip6tables - BOOLEAN
2785 - 1 : pass bridged IPv6 traffic to ip6tables' chains.
2790 bridge-nf-filter-vlan-tagged - BOOLEAN
2791 - 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2796 bridge-nf-filter-pppoe-tagged - BOOLEAN
2797 - 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2802 bridge-nf-pass-vlan-input-dev - BOOLEAN
2803 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2804 interface on the bridge and set the netfilter input device to the
2805 vlan. This allows use of e.g. "iptables -i br0.1" and makes the
2806 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no
2807 matching vlan interface is found, or this switch is off, the input
2808 device is set to the bridge interface.
2810 - 0: disable bridge netfilter vlan interface lookup.
2814 ``proc/sys/net/sctp/*`` Variables:
2815 ==================================
2817 addip_enable - BOOLEAN
2818 Enable or disable extension of Dynamic Address Reconfiguration
2819 (ADD-IP) functionality specified in RFC5061. This extension provides
2820 the ability to dynamically add and remove new addresses for the SCTP
2823 1: Enable extension.
2825 0: Disable extension.
2830 Enable or disable pf (pf is short for potentially failed) state. A value
2831 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2832 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2833 Since pf_retrans and path_max_retrans can be changed by userspace
2834 application, sometimes user expects to disable pf state by the value of
2835 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2836 or path_max_retrans is changed by the user application, this pf state is
2837 enabled. As such, it is necessary to add this to dynamically enable
2838 and disable pf state. See:
2839 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2849 Unset or enable/disable pf (pf is short for potentially failed) state
2850 exposure. Applications can control the exposure of the PF path state
2851 in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2852 sockopt. When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2853 SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2854 can be got via SCTP_GET_PEER_ADDR_INFO sockopt; When it's enabled,
2855 a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2856 SCTP_PF state and a SCTP_PF-state transport info can be got via
2857 SCTP_GET_PEER_ADDR_INFO sockopt; When it's disabled, no
2858 SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2859 trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2862 0: Unset pf state exposure, Compatible with old applications.
2864 1: Disable pf state exposure.
2866 2: Enable pf state exposure.
2870 addip_noauth_enable - BOOLEAN
2871 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2872 authentication to protect the operations of adding or removing new
2873 addresses. This requirement is mandated so that unauthorized hosts
2874 would not be able to hijack associations. However, older
2875 implementations may not have implemented this requirement while
2876 allowing the ADD-IP extension. For reasons of interoperability,
2877 we provide this variable to control the enforcement of the
2878 authentication requirement.
2880 == ===============================================================
2881 1 Allow ADD-IP extension to be used without authentication. This
2882 should only be set in a closed environment for interoperability
2883 with older implementations.
2885 0 Enforce the authentication requirement
2886 == ===============================================================
2890 auth_enable - BOOLEAN
2891 Enable or disable Authenticated Chunks extension. This extension
2892 provides the ability to send and receive authenticated chunks and is
2893 required for secure operation of Dynamic Address Reconfiguration
2896 - 1: Enable this extension.
2897 - 0: Disable this extension.
2901 prsctp_enable - BOOLEAN
2902 Enable or disable the Partial Reliability extension (RFC3758) which
2903 is used to notify peers that a given DATA should no longer be expected.
2905 - 1: Enable extension
2911 The limit of the number of new packets that can be initially sent. It
2912 controls how bursty the generated traffic can be.
2916 association_max_retrans - INTEGER
2917 Set the maximum number for retransmissions that an association can
2918 attempt deciding that the remote end is unreachable. If this value
2919 is exceeded, the association is terminated.
2923 max_init_retransmits - INTEGER
2924 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2925 that an association will attempt before declaring the destination
2926 unreachable and terminating.
2930 path_max_retrans - INTEGER
2931 The maximum number of retransmissions that will be attempted on a given
2932 path. Once this threshold is exceeded, the path is considered
2933 unreachable, and new traffic will use a different path when the
2934 association is multihomed.
2938 pf_retrans - INTEGER
2939 The number of retransmissions that will be attempted on a given path
2940 before traffic is redirected to an alternate transport (should one
2941 exist). Note this is distinct from path_max_retrans, as a path that
2942 passes the pf_retrans threshold can still be used. Its only
2943 deprioritized when a transmission path is selected by the stack. This
2944 setting is primarily used to enable fast failover mechanisms without
2945 having to reduce path_max_retrans to a very low value. See:
2946 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2947 for details. Note also that a value of pf_retrans > path_max_retrans
2948 disables this feature. Since both pf_retrans and path_max_retrans can
2949 be changed by userspace application, a variable pf_enable is used to
2954 ps_retrans - INTEGER
2955 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
2956 from section-5 "Primary Path Switchover" in rfc7829. The primary path
2957 will be changed to another active path when the path error counter on
2958 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
2959 to continue data transmission on a new working path even when the old
2960 primary destination address becomes active again". Note this feature
2961 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
2962 and its value can't be less than 'pf_retrans' when changing by sysctl.
2966 rto_initial - INTEGER
2967 The initial round trip timeout value in milliseconds that will be used
2968 in calculating round trip times. This is the initial time interval
2969 for retransmissions.
2974 The maximum value (in milliseconds) of the round trip timeout. This
2975 is the largest time interval that can elapse between retransmissions.
2980 The minimum value (in milliseconds) of the round trip timeout. This
2981 is the smallest time interval the can elapse between retransmissions.
2985 hb_interval - INTEGER
2986 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2987 are sent at the specified interval on idle paths to probe the state of
2988 a given path between 2 associations.
2992 sack_timeout - INTEGER
2993 The amount of time (in milliseconds) that the implementation will wait
2998 valid_cookie_life - INTEGER
2999 The default lifetime of the SCTP cookie (in milliseconds). The cookie
3000 is used during association establishment.
3004 cookie_preserve_enable - BOOLEAN
3005 Enable or disable the ability to extend the lifetime of the SCTP cookie
3006 that is used during the establishment phase of SCTP association
3008 - 1: Enable cookie lifetime extension.
3013 cookie_hmac_alg - STRING
3014 Select the hmac algorithm used when generating the cookie value sent by
3015 a listening sctp socket to a connecting client in the INIT-ACK chunk.
3022 Ability to assign md5 or sha1 as the selected alg is predicated on the
3023 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
3024 CONFIG_CRYPTO_SHA1).
3026 Default: Dependent on configuration. MD5 if available, else SHA1 if
3027 available, else none.
3029 rcvbuf_policy - INTEGER
3030 Determines if the receive buffer is attributed to the socket or to
3031 association. SCTP supports the capability to create multiple
3032 associations on a single socket. When using this capability, it is
3033 possible that a single stalled association that's buffering a lot
3034 of data may block other associations from delivering their data by
3035 consuming all of the receive buffer space. To work around this,
3036 the rcvbuf_policy could be set to attribute the receiver buffer space
3037 to each association instead of the socket. This prevents the described
3040 - 1: rcvbuf space is per association
3041 - 0: rcvbuf space is per socket
3045 sndbuf_policy - INTEGER
3046 Similar to rcvbuf_policy above, this applies to send buffer space.
3048 - 1: Send buffer is tracked per association
3049 - 0: Send buffer is tracked per socket.
3053 sctp_mem - vector of 3 INTEGERs: min, pressure, max
3054 Number of pages allowed for queueing by all SCTP sockets.
3056 min: Below this number of pages SCTP is not bothered about its
3057 memory appetite. When amount of memory allocated by SCTP exceeds
3058 this number, SCTP starts to moderate memory usage.
3060 pressure: This value was introduced to follow format of tcp_mem.
3062 max: Number of pages allowed for queueing by all SCTP sockets.
3064 Default is calculated at boot time from amount of available memory.
3066 sctp_rmem - vector of 3 INTEGERs: min, default, max
3067 Only the first value ("min") is used, "default" and "max" are
3070 min: Minimal size of receive buffer used by SCTP socket.
3071 It is guaranteed to each SCTP socket (but not association) even
3072 under moderate memory pressure.
3076 sctp_wmem - vector of 3 INTEGERs: min, default, max
3077 Only the first value ("min") is used, "default" and "max" are
3080 min: Minimum size of send buffer that can be used by SCTP sockets.
3081 It is guaranteed to each SCTP socket (but not association) even
3082 under moderate memory pressure.
3086 addr_scope_policy - INTEGER
3087 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
3089 - 0 - Disable IPv4 address scoping
3090 - 1 - Enable IPv4 address scoping
3091 - 2 - Follow draft but allow IPv4 private addresses
3092 - 3 - Follow draft but allow IPv4 link local addresses
3097 The listening port for the local UDP tunneling sock. Normally it's
3098 using the IANA-assigned UDP port number 9899 (sctp-tunneling).
3100 This UDP sock is used for processing the incoming UDP-encapsulated
3101 SCTP packets (from RFC6951), and shared by all applications in the
3102 same net namespace. This UDP sock will be closed when the value is
3105 The value will also be used to set the src port of the UDP header
3106 for the outgoing UDP-encapsulated SCTP packets. For the dest port,
3107 please refer to 'encap_port' below.
3111 encap_port - INTEGER
3112 The default remote UDP encapsulation port.
3114 This value is used to set the dest port of the UDP header for the
3115 outgoing UDP-encapsulated SCTP packets by default. Users can also
3116 change the value for each sock/asoc/transport by using setsockopt.
3117 For further information, please refer to RFC6951.
3119 Note that when connecting to a remote server, the client should set
3120 this to the port that the UDP tunneling sock on the peer server is
3121 listening to and the local UDP tunneling sock on the client also
3122 must be started. On the server, it would get the encap_port from
3123 the incoming packet's source port.
3127 plpmtud_probe_interval - INTEGER
3128 The time interval (in milliseconds) for the PLPMTUD probe timer,
3129 which is configured to expire after this period to receive an
3130 acknowledgment to a probe packet. This is also the time interval
3131 between the probes for the current pmtu when the probe search
3134 PLPMTUD will be disabled when 0 is set, and other values for it
3139 reconf_enable - BOOLEAN
3140 Enable or disable extension of Stream Reconfiguration functionality
3141 specified in RFC6525. This extension provides the ability to "reset"
3142 a stream, and it includes the Parameters of "Outgoing/Incoming SSN
3143 Reset", "SSN/TSN Reset" and "Add Outgoing/Incoming Streams".
3145 - 1: Enable extension.
3146 - 0: Disable extension.
3150 intl_enable - BOOLEAN
3151 Enable or disable extension of User Message Interleaving functionality
3152 specified in RFC8260. This extension allows the interleaving of user
3153 messages sent on different streams. With this feature enabled, I-DATA
3154 chunk will replace DATA chunk to carry user messages if also supported
3155 by the peer. Note that to use this feature, one needs to set this option
3156 to 1 and also needs to set socket options SCTP_FRAGMENT_INTERLEAVE to 2
3157 and SCTP_INTERLEAVING_SUPPORTED to 1.
3159 - 1: Enable extension.
3160 - 0: Disable extension.
3164 ecn_enable - BOOLEAN
3165 Control use of Explicit Congestion Notification (ECN) by SCTP.
3166 Like in TCP, ECN is used only when both ends of the SCTP connection
3167 indicate support for it. This feature is useful in avoiding losses
3168 due to congestion by allowing supporting routers to signal congestion
3169 before having to drop packets.
3176 l3mdev_accept - BOOLEAN
3177 Enabling this option allows a "global" bound socket to work
3178 across L3 master domains (e.g., VRFs) with packets capable of
3179 being received regardless of the L3 domain in which they
3180 originated. Only valid when the kernel was compiled with
3181 CONFIG_NET_L3_MASTER_DEV.
3183 Default: 1 (enabled)
3186 ``/proc/sys/net/core/*``
3187 ========================
3189 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
3192 ``/proc/sys/net/unix/*``
3193 ========================
3195 max_dgram_qlen - INTEGER
3196 The maximum length of dgram socket receive queue