1 /proc/sys/net/ipv4/* Variables:
7 Forward Packets between interfaces.
9 This variable is special, its change resets all configuration
10 parameters to their default state (RFC1122 for hosts, RFC1812
13 ip_default_ttl - INTEGER
14 Default value of TTL field (Time To Live) for outgoing (but not
15 forwarded) IP packets. Should be between 1 and 255 inclusive.
16 Default: 64 (as recommended by RFC1700)
18 ip_no_pmtu_disc - INTEGER
19 Disable Path MTU Discovery. If enabled in mode 1 and a
20 fragmentation-required ICMP is received, the PMTU to this
21 destination will be set to min_pmtu (see below). You will need
22 to raise min_pmtu to the smallest interface MTU on your system
23 manually if you want to avoid locally generated fragments.
25 In mode 2 incoming Path MTU Discovery messages will be
26 discarded. Outgoing frames are handled the same as in mode 1,
27 implicitly setting IP_PMTUDISC_DONT on every created socket.
29 Mode 3 is a hardened pmtu discover mode. The kernel will only
30 accept fragmentation-needed errors if the underlying protocol
31 can verify them besides a plain socket lookup. Current
32 protocols for which pmtu events will be honored are TCP, SCTP
33 and DCCP as they verify e.g. the sequence number or the
34 association. This mode should not be enabled globally but is
35 only intended to secure e.g. name servers in namespaces where
36 TCP path mtu must still work but path MTU information of other
37 protocols should be discarded. If enabled globally this mode
38 could break other protocols.
44 default 552 - minimum discovered Path MTU
46 ip_forward_use_pmtu - BOOLEAN
47 By default we don't trust protocol path MTUs while forwarding
48 because they could be easily forged and can lead to unwanted
49 fragmentation by the router.
50 You only need to enable this if you have user-space software
51 which tries to discover path mtus by itself and depends on the
52 kernel honoring this information. This is normally not the
59 fwmark_reflect - BOOLEAN
60 Controls the fwmark of kernel-generated IPv4 reply packets that are not
61 associated with a socket for example, TCP RSTs or ICMP echo replies).
62 If unset, these packets have a fwmark of zero. If set, they have the
63 fwmark of the packet they are replying to.
66 fib_multipath_use_neigh - BOOLEAN
67 Use status of existing neighbor entry when determining nexthop for
68 multipath routes. If disabled, neighbor information is not used and
69 packets could be directed to a failed nexthop. Only valid for kernels
70 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
76 fib_multipath_hash_policy - INTEGER
77 Controls which hash policy to use for multipath routes. Only valid
78 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
83 2 - Layer 3 or inner Layer 3 if present
85 fib_sync_mem - UNSIGNED INTEGER
86 Amount of dirty memory from fib entries that can be backlogged before
87 synchronize_rcu is forced.
88 Default: 512kB Minimum: 64kB Maximum: 64MB
90 ip_forward_update_priority - INTEGER
91 Whether to update SKB priority from "TOS" field in IPv4 header after it
92 is forwarded. The new SKB priority is mapped from TOS field value
93 according to an rt_tos2priority table (see e.g. man tc-prio).
94 Default: 1 (Update priority.)
96 0 - Do not update priority.
99 route/max_size - INTEGER
100 Maximum number of routes allowed in the kernel. Increase
101 this when using large numbers of interfaces and/or routes.
102 From linux kernel 3.6 onwards, this is deprecated for ipv4
103 as route cache is no longer used.
105 neigh/default/gc_thresh1 - INTEGER
106 Minimum number of entries to keep. Garbage collector will not
107 purge entries if there are fewer than this number.
110 neigh/default/gc_thresh2 - INTEGER
111 Threshold when garbage collector becomes more aggressive about
112 purging entries. Entries older than 5 seconds will be cleared
113 when over this number.
116 neigh/default/gc_thresh3 - INTEGER
117 Maximum number of non-PERMANENT neighbor entries allowed. Increase
118 this when using large numbers of interfaces and when communicating
119 with large numbers of directly-connected peers.
122 neigh/default/unres_qlen_bytes - INTEGER
123 The maximum number of bytes which may be used by packets
124 queued for each unresolved address by other network layers.
126 Setting negative value is meaningless and will return error.
127 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
128 Exact value depends on architecture and kernel options,
129 but should be enough to allow queuing 256 packets
132 neigh/default/unres_qlen - INTEGER
133 The maximum number of packets which may be queued for each
134 unresolved address by other network layers.
135 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
136 Prior to linux 3.3, the default value is 3 which may cause
137 unexpected packet loss. The current default value is calculated
138 according to default value of unres_qlen_bytes and true size of
142 mtu_expires - INTEGER
143 Time, in seconds, that cached PMTU information is kept.
145 min_adv_mss - INTEGER
146 The advertised MSS depends on the first hop route MTU, but will
147 never be lower than this setting.
151 ipfrag_high_thresh - LONG INTEGER
152 Maximum memory used to reassemble IP fragments.
154 ipfrag_low_thresh - LONG INTEGER
155 (Obsolete since linux-4.17)
156 Maximum memory used to reassemble IP fragments before the kernel
157 begins to remove incomplete fragment queues to free up resources.
158 The kernel still accepts new fragments for defragmentation.
160 ipfrag_time - INTEGER
161 Time in seconds to keep an IP fragment in memory.
163 ipfrag_max_dist - INTEGER
164 ipfrag_max_dist is a non-negative integer value which defines the
165 maximum "disorder" which is allowed among fragments which share a
166 common IP source address. Note that reordering of packets is
167 not unusual, but if a large number of fragments arrive from a source
168 IP address while a particular fragment queue remains incomplete, it
169 probably indicates that one or more fragments belonging to that queue
170 have been lost. When ipfrag_max_dist is positive, an additional check
171 is done on fragments before they are added to a reassembly queue - if
172 ipfrag_max_dist (or more) fragments have arrived from a particular IP
173 address between additions to any IP fragment queue using that source
174 address, it's presumed that one or more fragments in the queue are
175 lost. The existing fragment queue will be dropped, and a new one
176 started. An ipfrag_max_dist value of zero disables this check.
178 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
179 result in unnecessarily dropping fragment queues when normal
180 reordering of packets occurs, which could lead to poor application
181 performance. Using a very large value, e.g. 50000, increases the
182 likelihood of incorrectly reassembling IP fragments that originate
183 from different IP datagrams, which could result in data corruption.
188 inet_peer_threshold - INTEGER
189 The approximate size of the storage. Starting from this threshold
190 entries will be thrown aggressively. This threshold also determines
191 entries' time-to-live and time intervals between garbage collection
192 passes. More entries, less time-to-live, less GC interval.
194 inet_peer_minttl - INTEGER
195 Minimum time-to-live of entries. Should be enough to cover fragment
196 time-to-live on the reassembling side. This minimum time-to-live is
197 guaranteed if the pool size is less than inet_peer_threshold.
200 inet_peer_maxttl - INTEGER
201 Maximum time-to-live of entries. Unused entries will expire after
202 this period of time if there is no memory pressure on the pool (i.e.
203 when the number of entries in the pool is very small).
209 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
210 Defaults to 4096. (Was 128 before linux-5.4)
211 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
213 tcp_abort_on_overflow - BOOLEAN
214 If listening service is too slow to accept new connections,
215 reset them. Default state is FALSE. It means that if overflow
216 occurred due to a burst, connection will recover. Enable this
217 option _only_ if you are really sure that listening daemon
218 cannot be tuned to accept connections faster. Enabling this
219 option can harm clients of your server.
221 tcp_adv_win_scale - INTEGER
222 Count buffering overhead as bytes/2^tcp_adv_win_scale
223 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
225 Possible values are [-31, 31], inclusive.
228 tcp_allowed_congestion_control - STRING
229 Show/set the congestion control choices available to non-privileged
230 processes. The list is a subset of those listed in
231 tcp_available_congestion_control.
232 Default is "reno" and the default setting (tcp_congestion_control).
234 tcp_app_win - INTEGER
235 Reserve max(window/2^tcp_app_win, mss) of window for application
236 buffer. Value 0 is special, it means that nothing is reserved.
239 tcp_autocorking - BOOLEAN
240 Enable TCP auto corking :
241 When applications do consecutive small write()/sendmsg() system calls,
242 we try to coalesce these small writes as much as possible, to lower
243 total amount of sent packets. This is done if at least one prior
244 packet for the flow is waiting in Qdisc queues or device transmit
245 queue. Applications can still use TCP_CORK for optimal behavior
246 when they know how/when to uncork their sockets.
249 tcp_available_congestion_control - STRING
250 Shows the available congestion control choices that are registered.
251 More congestion control algorithms may be available as modules,
254 tcp_base_mss - INTEGER
255 The initial value of search_low to be used by the packetization layer
256 Path MTU discovery (MTU probing). If MTU probing is enabled,
257 this is the initial MSS used by the connection.
259 tcp_mtu_probe_floor - INTEGER
260 If MTU probing is enabled this caps the minimum MSS used for search_low
265 tcp_min_snd_mss - INTEGER
266 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
267 as described in RFC 1122 and RFC 6691.
268 If this ADVMSS option is smaller than tcp_min_snd_mss,
269 it is silently capped to tcp_min_snd_mss.
271 Default : 48 (at least 8 bytes of payload per segment)
273 tcp_congestion_control - STRING
274 Set the congestion control algorithm to be used for new
275 connections. The algorithm "reno" is always available, but
276 additional choices may be available based on kernel configuration.
277 Default is set as part of kernel configuration.
278 For passive connections, the listener congestion control choice
280 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
283 Allows TCP to send "duplicate" SACKs.
285 tcp_early_retrans - INTEGER
286 Tail loss probe (TLP) converts RTOs occurring due to tail
287 losses into fast recovery (draft-ietf-tcpm-rack). Note that
288 TLP requires RACK to function properly (see tcp_recovery below)
295 Control use of Explicit Congestion Notification (ECN) by TCP.
296 ECN is used only when both ends of the TCP connection indicate
297 support for it. This feature is useful in avoiding losses due
298 to congestion by allowing supporting routers to signal
299 congestion before having to drop packets.
301 0 Disable ECN. Neither initiate nor accept ECN.
302 1 Enable ECN when requested by incoming connections and
303 also request ECN on outgoing connection attempts.
304 2 Enable ECN when requested by incoming connections
305 but do not request ECN on outgoing connections.
308 tcp_ecn_fallback - BOOLEAN
309 If the kernel detects that ECN connection misbehaves, enable fall
310 back to non-ECN. Currently, this knob implements the fallback
311 from RFC3168, section 6.1.1.1., but we reserve that in future,
312 additional detection mechanisms could be implemented under this
313 knob. The value is not used, if tcp_ecn or per route (or congestion
314 control) ECN settings are disabled.
315 Default: 1 (fallback enabled)
318 This is a legacy option, it has no effect anymore.
320 tcp_fin_timeout - INTEGER
321 The length of time an orphaned (no longer referenced by any
322 application) connection will remain in the FIN_WAIT_2 state
323 before it is aborted at the local end. While a perfectly
324 valid "receive only" state for an un-orphaned connection, an
325 orphaned connection in FIN_WAIT_2 state could otherwise wait
326 forever for the remote to close its end of the connection.
331 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
332 F-RTO is an enhanced recovery algorithm for TCP retransmission
333 timeouts. It is particularly beneficial in networks where the
334 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
335 modification. It does not require any support from the peer.
337 By default it's enabled with a non-zero value. 0 disables F-RTO.
339 tcp_fwmark_accept - BOOLEAN
340 If set, incoming connections to listening sockets that do not have a
341 socket mark will set the mark of the accepting socket to the fwmark of
342 the incoming SYN packet. This will cause all packets on that connection
343 (starting from the first SYNACK) to be sent with that fwmark. The
344 listening socket's mark is unchanged. Listening sockets that already
345 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
350 tcp_invalid_ratelimit - INTEGER
351 Limit the maximal rate for sending duplicate acknowledgments
352 in response to incoming TCP packets that are for an existing
353 connection but that are invalid due to any of these reasons:
355 (a) out-of-window sequence number,
356 (b) out-of-window acknowledgment number, or
357 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
359 This can help mitigate simple "ack loop" DoS attacks, wherein
360 a buggy or malicious middlebox or man-in-the-middle can
361 rewrite TCP header fields in manner that causes each endpoint
362 to think that the other is sending invalid TCP segments, thus
363 causing each side to send an unterminating stream of duplicate
364 acknowledgments for invalid segments.
366 Using 0 disables rate-limiting of dupacks in response to
367 invalid segments; otherwise this value specifies the minimal
368 space between sending such dupacks, in milliseconds.
370 Default: 500 (milliseconds).
372 tcp_keepalive_time - INTEGER
373 How often TCP sends out keepalive messages when keepalive is enabled.
376 tcp_keepalive_probes - INTEGER
377 How many keepalive probes TCP sends out, until it decides that the
378 connection is broken. Default value: 9.
380 tcp_keepalive_intvl - INTEGER
381 How frequently the probes are send out. Multiplied by
382 tcp_keepalive_probes it is time to kill not responding connection,
383 after probes started. Default value: 75sec i.e. connection
384 will be aborted after ~11 minutes of retries.
386 tcp_l3mdev_accept - BOOLEAN
387 Enables child sockets to inherit the L3 master device index.
388 Enabling this option allows a "global" listen socket to work
389 across L3 master domains (e.g., VRFs) with connected sockets
390 derived from the listen socket to be bound to the L3 domain in
391 which the packets originated. Only valid when the kernel was
392 compiled with CONFIG_NET_L3_MASTER_DEV.
393 Default: 0 (disabled)
395 tcp_low_latency - BOOLEAN
396 This is a legacy option, it has no effect anymore.
398 tcp_max_orphans - INTEGER
399 Maximal number of TCP sockets not attached to any user file handle,
400 held by system. If this number is exceeded orphaned connections are
401 reset immediately and warning is printed. This limit exists
402 only to prevent simple DoS attacks, you _must_ not rely on this
403 or lower the limit artificially, but rather increase it
404 (probably, after increasing installed memory),
405 if network conditions require more than default value,
406 and tune network services to linger and kill such states
407 more aggressively. Let me to remind again: each orphan eats
408 up to ~64K of unswappable memory.
410 tcp_max_syn_backlog - INTEGER
411 Maximal number of remembered connection requests (SYN_RECV),
412 which have not received an acknowledgment from connecting client.
413 This is a per-listener limit.
414 The minimal value is 128 for low memory machines, and it will
415 increase in proportion to the memory of machine.
416 If server suffers from overload, try increasing this number.
417 Remember to also check /proc/sys/net/core/somaxconn
418 A SYN_RECV request socket consumes about 304 bytes of memory.
420 tcp_max_tw_buckets - INTEGER
421 Maximal number of timewait sockets held by system simultaneously.
422 If this number is exceeded time-wait socket is immediately destroyed
423 and warning is printed. This limit exists only to prevent
424 simple DoS attacks, you _must_ not lower the limit artificially,
425 but rather increase it (probably, after increasing installed memory),
426 if network conditions require more than default value.
428 tcp_mem - vector of 3 INTEGERs: min, pressure, max
429 min: below this number of pages TCP is not bothered about its
432 pressure: when amount of memory allocated by TCP exceeds this number
433 of pages, TCP moderates its memory consumption and enters memory
434 pressure mode, which is exited when memory consumption falls
437 max: number of pages allowed for queueing by all TCP sockets.
439 Defaults are calculated at boot time from amount of available
442 tcp_min_rtt_wlen - INTEGER
443 The window length of the windowed min filter to track the minimum RTT.
444 A shorter window lets a flow more quickly pick up new (higher)
445 minimum RTT when it is moved to a longer path (e.g., due to traffic
446 engineering). A longer window makes the filter more resistant to RTT
447 inflations such as transient congestion. The unit is seconds.
448 Possible values: 0 - 86400 (1 day)
451 tcp_moderate_rcvbuf - BOOLEAN
452 If set, TCP performs receive buffer auto-tuning, attempting to
453 automatically size the buffer (no greater than tcp_rmem[2]) to
454 match the size required by the path for full throughput. Enabled by
457 tcp_mtu_probing - INTEGER
458 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
461 1 - Disabled by default, enabled when an ICMP black hole detected
462 2 - Always enabled, use initial MSS of tcp_base_mss.
464 tcp_probe_interval - UNSIGNED INTEGER
465 Controls how often to start TCP Packetization-Layer Path MTU
466 Discovery reprobe. The default is reprobing every 10 minutes as
469 tcp_probe_threshold - INTEGER
470 Controls when TCP Packetization-Layer Path MTU Discovery probing
471 will stop in respect to the width of search range in bytes. Default
474 tcp_no_metrics_save - BOOLEAN
475 By default, TCP saves various connection metrics in the route cache
476 when the connection closes, so that connections established in the
477 near future can use these to set initial conditions. Usually, this
478 increases overall performance, but may sometimes cause performance
479 degradation. If set, TCP will not cache metrics on closing
482 tcp_orphan_retries - INTEGER
483 This value influences the timeout of a locally closed TCP connection,
484 when RTO retransmissions remain unacknowledged.
485 See tcp_retries2 for more details.
487 The default value is 8.
488 If your machine is a loaded WEB server,
489 you should think about lowering this value, such sockets
490 may consume significant resources. Cf. tcp_max_orphans.
492 tcp_recovery - INTEGER
493 This value is a bitmap to enable various experimental loss recovery
496 RACK: 0x1 enables the RACK loss detection for fast detection of lost
497 retransmissions and tail drops. It also subsumes and disables
498 RFC6675 recovery for SACK connections.
499 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
500 RACK: 0x4 disables RACK's DUPACK threshold heuristic
504 tcp_reordering - INTEGER
505 Initial reordering level of packets in a TCP stream.
506 TCP stack can then dynamically adjust flow reordering level
507 between this initial value and tcp_max_reordering
510 tcp_max_reordering - INTEGER
511 Maximal reordering level of packets in a TCP stream.
512 300 is a fairly conservative value, but you might increase it
513 if paths are using per packet load balancing (like bonding rr mode)
516 tcp_retrans_collapse - BOOLEAN
517 Bug-to-bug compatibility with some broken printers.
518 On retransmit try to send bigger packets to work around bugs in
521 tcp_retries1 - INTEGER
522 This value influences the time, after which TCP decides, that
523 something is wrong due to unacknowledged RTO retransmissions,
524 and reports this suspicion to the network layer.
525 See tcp_retries2 for more details.
527 RFC 1122 recommends at least 3 retransmissions, which is the
530 tcp_retries2 - INTEGER
531 This value influences the timeout of an alive TCP connection,
532 when RTO retransmissions remain unacknowledged.
533 Given a value of N, a hypothetical TCP connection following
534 exponential backoff with an initial RTO of TCP_RTO_MIN would
535 retransmit N times before killing the connection at the (N+1)th RTO.
537 The default value of 15 yields a hypothetical timeout of 924.6
538 seconds and is a lower bound for the effective timeout.
539 TCP will effectively time out at the first RTO which exceeds the
540 hypothetical timeout.
542 RFC 1122 recommends at least 100 seconds for the timeout,
543 which corresponds to a value of at least 8.
545 tcp_rfc1337 - BOOLEAN
546 If set, the TCP stack behaves conforming to RFC1337. If unset,
547 we are not conforming to RFC, but prevent TCP TIME_WAIT
551 tcp_rmem - vector of 3 INTEGERs: min, default, max
552 min: Minimal size of receive buffer used by TCP sockets.
553 It is guaranteed to each TCP socket, even under moderate memory
557 default: initial size of receive buffer used by TCP sockets.
558 This value overrides net.core.rmem_default used by other protocols.
559 Default: 87380 bytes. This value results in window of 65535 with
560 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
561 less for default tcp_app_win. See below about these variables.
563 max: maximal size of receive buffer allowed for automatically
564 selected receiver buffers for TCP socket. This value does not override
565 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
566 automatic tuning of that socket's receive buffer size, in which
567 case this value is ignored.
568 Default: between 87380B and 6MB, depending on RAM size.
571 Enable select acknowledgments (SACKS).
573 tcp_comp_sack_delay_ns - LONG INTEGER
574 TCP tries to reduce number of SACK sent, using a timer
575 based on 5% of SRTT, capped by this sysctl, in nano seconds.
576 The default is 1ms, based on TSO autosizing period.
578 Default : 1,000,000 ns (1 ms)
580 tcp_comp_sack_nr - INTEGER
581 Max number of SACK that can be compressed.
582 Using 0 disables SACK compression.
586 tcp_slow_start_after_idle - BOOLEAN
587 If set, provide RFC2861 behavior and time out the congestion
588 window after an idle period. An idle period is defined at
589 the current RTO. If unset, the congestion window will not
590 be timed out after an idle period.
594 Use the Host requirements interpretation of the TCP urgent pointer field.
595 Most hosts use the older BSD interpretation, so if you turn this on
596 Linux might not communicate correctly with them.
599 tcp_synack_retries - INTEGER
600 Number of times SYNACKs for a passive TCP connection attempt will
601 be retransmitted. Should not be higher than 255. Default value
602 is 5, which corresponds to 31seconds till the last retransmission
603 with the current initial RTO of 1second. With this the final timeout
604 for a passive TCP connection will happen after 63seconds.
606 tcp_syncookies - BOOLEAN
607 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
608 Send out syncookies when the syn backlog queue of a socket
609 overflows. This is to prevent against the common 'SYN flood attack'
612 Note, that syncookies is fallback facility.
613 It MUST NOT be used to help highly loaded servers to stand
614 against legal connection rate. If you see SYN flood warnings
615 in your logs, but investigation shows that they occur
616 because of overload with legal connections, you should tune
617 another parameters until this warning disappear.
618 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
620 syncookies seriously violate TCP protocol, do not allow
621 to use TCP extensions, can result in serious degradation
622 of some services (f.e. SMTP relaying), visible not by you,
623 but your clients and relays, contacting you. While you see
624 SYN flood warnings in logs not being really flooded, your server
625 is seriously misconfigured.
627 If you want to test which effects syncookies have to your
628 network connections you can set this knob to 2 to enable
629 unconditionally generation of syncookies.
631 tcp_fastopen - INTEGER
632 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
635 The client support is enabled by flag 0x1 (on by default). The client
636 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
637 rather than connect() to send data in SYN.
639 The server support is enabled by flag 0x2 (off by default). Then
640 either enable for all listeners with another flag (0x400) or
641 enable individual listeners via TCP_FASTOPEN socket option with
642 the option value being the length of the syn-data backlog.
644 The values (bitmap) are
645 0x1: (client) enables sending data in the opening SYN on the client.
646 0x2: (server) enables the server support, i.e., allowing data in
647 a SYN packet to be accepted and passed to the
648 application before 3-way handshake finishes.
649 0x4: (client) send data in the opening SYN regardless of cookie
650 availability and without a cookie option.
651 0x200: (server) accept data-in-SYN w/o any cookie option present.
652 0x400: (server) enable all listeners to support Fast Open by
653 default without explicit TCP_FASTOPEN socket option.
657 Note that that additional client or server features are only
658 effective if the basic support (0x1 and 0x2) are enabled respectively.
660 tcp_fastopen_blackhole_timeout_sec - INTEGER
661 Initial time period in second to disable Fastopen on active TCP sockets
662 when a TFO firewall blackhole issue happens.
663 This time period will grow exponentially when more blackhole issues
664 get detected right after Fastopen is re-enabled and will reset to
665 initial value when the blackhole issue goes away.
666 0 to disable the blackhole detection.
667 By default, it is set to 1hr.
669 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
670 The list consists of a primary key and an optional backup key. The
671 primary key is used for both creating and validating cookies, while the
672 optional backup key is only used for validating cookies. The purpose of
673 the backup key is to maximize TFO validation when keys are rotated.
675 A randomly chosen primary key may be configured by the kernel if
676 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
677 TCP_FASTOPEN setsockopt() optname is set and a key has not been
678 previously configured via sysctl. If keys are configured via
679 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
680 per-socket keys will be used instead of any keys that are specified via
683 A key is specified as 4 8-digit hexadecimal integers which are separated
684 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
685 omitted. A primary and a backup key may be specified by separating them
686 by a comma. If only one key is specified, it becomes the primary key and
687 any previously configured backup keys are removed.
689 tcp_syn_retries - INTEGER
690 Number of times initial SYNs for an active TCP connection attempt
691 will be retransmitted. Should not be higher than 127. Default value
692 is 6, which corresponds to 63seconds till the last retransmission
693 with the current initial RTO of 1second. With this the final timeout
694 for an active TCP connection attempt will happen after 127seconds.
696 tcp_timestamps - INTEGER
697 Enable timestamps as defined in RFC1323.
699 1: Enable timestamps as defined in RFC1323 and use random offset for
700 each connection rather than only using the current time.
701 2: Like 1, but without random offsets.
704 tcp_min_tso_segs - INTEGER
705 Minimal number of segments per TSO frame.
706 Since linux-3.12, TCP does an automatic sizing of TSO frames,
707 depending on flow rate, instead of filling 64Kbytes packets.
708 For specific usages, it's possible to force TCP to build big
709 TSO frames. Note that TCP stack might split too big TSO packets
710 if available window is too small.
713 tcp_pacing_ss_ratio - INTEGER
714 sk->sk_pacing_rate is set by TCP stack using a ratio applied
715 to current rate. (current_rate = cwnd * mss / srtt)
716 If TCP is in slow start, tcp_pacing_ss_ratio is applied
717 to let TCP probe for bigger speeds, assuming cwnd can be
718 doubled every other RTT.
721 tcp_pacing_ca_ratio - INTEGER
722 sk->sk_pacing_rate is set by TCP stack using a ratio applied
723 to current rate. (current_rate = cwnd * mss / srtt)
724 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
725 is applied to conservatively probe for bigger throughput.
728 tcp_tso_win_divisor - INTEGER
729 This allows control over what percentage of the congestion window
730 can be consumed by a single TSO frame.
731 The setting of this parameter is a choice between burstiness and
732 building larger TSO frames.
735 tcp_tw_reuse - INTEGER
736 Enable reuse of TIME-WAIT sockets for new connections when it is
737 safe from protocol viewpoint.
740 2 - enable for loopback traffic only
741 It should not be changed without advice/request of technical
745 tcp_window_scaling - BOOLEAN
746 Enable window scaling as defined in RFC1323.
748 tcp_wmem - vector of 3 INTEGERs: min, default, max
749 min: Amount of memory reserved for send buffers for TCP sockets.
750 Each TCP socket has rights to use it due to fact of its birth.
753 default: initial size of send buffer used by TCP sockets. This
754 value overrides net.core.wmem_default used by other protocols.
755 It is usually lower than net.core.wmem_default.
758 max: Maximal amount of memory allowed for automatically tuned
759 send buffers for TCP sockets. This value does not override
760 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
761 automatic tuning of that socket's send buffer size, in which case
762 this value is ignored.
763 Default: between 64K and 4MB, depending on RAM size.
765 tcp_notsent_lowat - UNSIGNED INTEGER
766 A TCP socket can control the amount of unsent bytes in its write queue,
767 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
768 reports POLLOUT events if the amount of unsent bytes is below a per
769 socket value, and if the write queue is not full. sendmsg() will
770 also not add new buffers if the limit is hit.
772 This global variable controls the amount of unsent data for
773 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
774 to the global variable has immediate effect.
776 Default: UINT_MAX (0xFFFFFFFF)
778 tcp_workaround_signed_windows - BOOLEAN
779 If set, assume no receipt of a window scaling option means the
780 remote TCP is broken and treats the window as a signed quantity.
781 If unset, assume the remote TCP is not broken even if we do
782 not receive a window scaling option from them.
785 tcp_thin_linear_timeouts - BOOLEAN
786 Enable dynamic triggering of linear timeouts for thin streams.
787 If set, a check is performed upon retransmission by timeout to
788 determine if the stream is thin (less than 4 packets in flight).
789 As long as the stream is found to be thin, up to 6 linear
790 timeouts may be performed before exponential backoff mode is
791 initiated. This improves retransmission latency for
792 non-aggressive thin streams, often found to be time-dependent.
793 For more information on thin streams, see
794 Documentation/networking/tcp-thin.txt
797 tcp_limit_output_bytes - INTEGER
798 Controls TCP Small Queue limit per tcp socket.
799 TCP bulk sender tends to increase packets in flight until it
800 gets losses notifications. With SNDBUF autotuning, this can
801 result in a large amount of packets queued on the local machine
802 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
803 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
804 limits the number of bytes on qdisc or device to reduce artificial
805 RTT/cwnd and reduce bufferbloat.
806 Default: 1048576 (16 * 65536)
808 tcp_challenge_ack_limit - INTEGER
809 Limits number of Challenge ACK sent per second, as recommended
810 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
813 tcp_rx_skb_cache - BOOLEAN
814 Controls a per TCP socket cache of one skb, that might help
815 performance of some workloads. This might be dangerous
816 on systems with a lot of TCP sockets, since it increases
819 Default: 0 (disabled)
823 udp_l3mdev_accept - BOOLEAN
824 Enabling this option allows a "global" bound socket to work
825 across L3 master domains (e.g., VRFs) with packets capable of
826 being received regardless of the L3 domain in which they
827 originated. Only valid when the kernel was compiled with
828 CONFIG_NET_L3_MASTER_DEV.
829 Default: 0 (disabled)
831 udp_mem - vector of 3 INTEGERs: min, pressure, max
832 Number of pages allowed for queueing by all UDP sockets.
834 min: Below this number of pages UDP is not bothered about its
835 memory appetite. When amount of memory allocated by UDP exceeds
836 this number, UDP starts to moderate memory usage.
838 pressure: This value was introduced to follow format of tcp_mem.
840 max: Number of pages allowed for queueing by all UDP sockets.
842 Default is calculated at boot time from amount of available memory.
844 udp_rmem_min - INTEGER
845 Minimal size of receive buffer used by UDP sockets in moderation.
846 Each UDP socket is able to use the size for receiving data, even if
847 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
850 udp_wmem_min - INTEGER
851 Minimal size of send buffer used by UDP sockets in moderation.
852 Each UDP socket is able to use the size for sending data, even if
853 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
858 raw_l3mdev_accept - BOOLEAN
859 Enabling this option allows a "global" bound socket to work
860 across L3 master domains (e.g., VRFs) with packets capable of
861 being received regardless of the L3 domain in which they
862 originated. Only valid when the kernel was compiled with
863 CONFIG_NET_L3_MASTER_DEV.
868 cipso_cache_enable - BOOLEAN
869 If set, enable additions to and lookups from the CIPSO label mapping
870 cache. If unset, additions are ignored and lookups always result in a
871 miss. However, regardless of the setting the cache is still
872 invalidated when required when means you can safely toggle this on and
873 off and the cache will always be "safe".
876 cipso_cache_bucket_size - INTEGER
877 The CIPSO label cache consists of a fixed size hash table with each
878 hash bucket containing a number of cache entries. This variable limits
879 the number of entries in each hash bucket; the larger the value the
880 more CIPSO label mappings that can be cached. When the number of
881 entries in a given hash bucket reaches this limit adding new entries
882 causes the oldest entry in the bucket to be removed to make room.
885 cipso_rbm_optfmt - BOOLEAN
886 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
887 the CIPSO draft specification (see Documentation/netlabel for details).
888 This means that when set the CIPSO tag will be padded with empty
889 categories in order to make the packet data 32-bit aligned.
892 cipso_rbm_structvalid - BOOLEAN
893 If set, do a very strict check of the CIPSO option when
894 ip_options_compile() is called. If unset, relax the checks done during
895 ip_options_compile(). Either way is "safe" as errors are caught else
896 where in the CIPSO processing code but setting this to 0 (False) should
897 result in less work (i.e. it should be faster) but could cause problems
898 with other implementations that require strict checking.
903 ip_local_port_range - 2 INTEGERS
904 Defines the local port range that is used by TCP and UDP to
905 choose the local port. The first number is the first, the
906 second the last local port number.
907 If possible, it is better these numbers have different parity.
908 (one even and one odd values)
909 The default values are 32768 and 60999 respectively.
911 ip_local_reserved_ports - list of comma separated ranges
912 Specify the ports which are reserved for known third-party
913 applications. These ports will not be used by automatic port
914 assignments (e.g. when calling connect() or bind() with port
915 number 0). Explicit port allocation behavior is unchanged.
917 The format used for both input and output is a comma separated
918 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
919 10). Writing to the file will clear all previously reserved
920 ports and update the current list with the one given in the
923 Note that ip_local_port_range and ip_local_reserved_ports
924 settings are independent and both are considered by the kernel
925 when determining which ports are available for automatic port
928 You can reserve ports which are not in the current
929 ip_local_port_range, e.g.:
931 $ cat /proc/sys/net/ipv4/ip_local_port_range
933 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
936 although this is redundant. However such a setting is useful
937 if later the port range is changed to a value that will
938 include the reserved ports.
942 ip_unprivileged_port_start - INTEGER
943 This is a per-namespace sysctl. It defines the first
944 unprivileged port in the network namespace. Privileged ports
945 require root or CAP_NET_BIND_SERVICE in order to bind to them.
946 To disable all privileged ports, set this to 0. It may not
947 overlap with the ip_local_reserved_ports range.
951 ip_nonlocal_bind - BOOLEAN
952 If set, allows processes to bind() to non-local IP addresses,
953 which can be quite useful - but may break some applications.
957 If set non-zero, enables support for dynamic addresses.
958 If set to a non-zero value larger than 1, a kernel log
959 message will be printed when dynamic address rewriting
963 ip_early_demux - BOOLEAN
964 Optimize input packet processing down to one demux for
965 certain kinds of local sockets. Currently we only do this
966 for established TCP and connected UDP sockets.
968 It may add an additional cost for pure routing workloads that
969 reduces overall throughput, in such case you should disable it.
972 tcp_early_demux - BOOLEAN
973 Enable early demux for established TCP sockets.
976 udp_early_demux - BOOLEAN
977 Enable early demux for connected UDP sockets. Disable this if
978 your system could experience more unconnected load.
981 icmp_echo_ignore_all - BOOLEAN
982 If set non-zero, then the kernel will ignore all ICMP ECHO
986 icmp_echo_ignore_broadcasts - BOOLEAN
987 If set non-zero, then the kernel will ignore all ICMP ECHO and
988 TIMESTAMP requests sent to it via broadcast/multicast.
991 icmp_ratelimit - INTEGER
992 Limit the maximal rates for sending ICMP packets whose type matches
993 icmp_ratemask (see below) to specific targets.
994 0 to disable any limiting,
995 otherwise the minimal space between responses in milliseconds.
996 Note that another sysctl, icmp_msgs_per_sec limits the number
997 of ICMP packets sent on all targets.
1000 icmp_msgs_per_sec - INTEGER
1001 Limit maximal number of ICMP packets sent per second from this host.
1002 Only messages whose type matches icmp_ratemask (see below) are
1003 controlled by this limit.
1006 icmp_msgs_burst - INTEGER
1007 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1008 while icmp_msgs_burst controls the burst size of these packets.
1011 icmp_ratemask - INTEGER
1012 Mask made of ICMP types for which rates are being limited.
1013 Significant bits: IHGFEDCBA9876543210
1014 Default mask: 0000001100000011000 (6168)
1016 Bit definitions (see include/linux/icmp.h):
1018 3 Destination Unreachable *
1023 C Parameter Problem *
1028 H Address Mask Request
1029 I Address Mask Reply
1031 * These are rate limited by default (see default mask above)
1033 icmp_ignore_bogus_error_responses - BOOLEAN
1034 Some routers violate RFC1122 by sending bogus responses to broadcast
1035 frames. Such violations are normally logged via a kernel warning.
1036 If this is set to TRUE, the kernel will not give such warnings, which
1037 will avoid log file clutter.
1040 icmp_errors_use_inbound_ifaddr - BOOLEAN
1042 If zero, icmp error messages are sent with the primary address of
1043 the exiting interface.
1045 If non-zero, the message will be sent with the primary address of
1046 the interface that received the packet that caused the icmp error.
1047 This is the behaviour network many administrators will expect from
1048 a router. And it can make debugging complicated network layouts
1051 Note that if no primary address exists for the interface selected,
1052 then the primary address of the first non-loopback interface that
1053 has one will be used regardless of this setting.
1057 igmp_max_memberships - INTEGER
1058 Change the maximum number of multicast groups we can subscribe to.
1061 Theoretical maximum value is bounded by having to send a membership
1062 report in a single datagram (i.e. the report can't span multiple
1063 datagrams, or risk confusing the switch and leaving groups you don't
1066 The number of supported groups 'M' is bounded by the number of group
1067 report entries you can fit into a single datagram of 65535 bytes.
1069 M = 65536-sizeof (ip header)/(sizeof(Group record))
1071 Group records are variable length, with a minimum of 12 bytes.
1072 So net.ipv4.igmp_max_memberships should not be set higher than:
1074 (65536-24) / 12 = 5459
1076 The value 5459 assumes no IP header options, so in practice
1077 this number may be lower.
1079 igmp_max_msf - INTEGER
1080 Maximum number of addresses allowed in the source filter list for a
1085 Controls the IGMP query robustness variable (see RFC2236 8.1).
1086 Default: 2 (as specified by RFC2236 8.1)
1087 Minimum: 1 (as specified by RFC6636 4.5)
1089 force_igmp_version - INTEGER
1090 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1091 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1092 Present timer expires.
1093 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1094 receive IGMPv2/v3 query.
1095 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1096 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1097 3 - Enforce to use IGMP version 3. The same react with default 0.
1099 Note: this is not the same with force_mld_version because IGMPv3 RFC3376
1100 Security Considerations does not have clear description that we could
1101 ignore other version messages completely as MLDv2 RFC3810. So make
1102 this value as default 0 is recommended.
1104 conf/interface/* changes special settings per interface (where
1105 "interface" is the name of your network interface)
1107 conf/all/* is special, changes the settings for all interfaces
1109 log_martians - BOOLEAN
1110 Log packets with impossible addresses to kernel log.
1111 log_martians for the interface will be enabled if at least one of
1112 conf/{all,interface}/log_martians is set to TRUE,
1113 it will be disabled otherwise
1115 accept_redirects - BOOLEAN
1116 Accept ICMP redirect messages.
1117 accept_redirects for the interface will be enabled if:
1118 - both conf/{all,interface}/accept_redirects are TRUE in the case
1119 forwarding for the interface is enabled
1121 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1122 case forwarding for the interface is disabled
1123 accept_redirects for the interface will be disabled otherwise
1127 forwarding - BOOLEAN
1128 Enable IP forwarding on this interface. This controls whether packets
1129 received _on_ this interface can be forwarded.
1131 mc_forwarding - BOOLEAN
1132 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1133 and a multicast routing daemon is required.
1134 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1135 routing for the interface
1138 Integer value used to differentiate the devices by the medium they
1139 are attached to. Two devices can have different id values when
1140 the broadcast packets are received only on one of them.
1141 The default value 0 means that the device is the only interface
1142 to its medium, value of -1 means that medium is not known.
1144 Currently, it is used to change the proxy_arp behavior:
1145 the proxy_arp feature is enabled for packets forwarded between
1146 two devices attached to different media.
1150 proxy_arp for the interface will be enabled if at least one of
1151 conf/{all,interface}/proxy_arp is set to TRUE,
1152 it will be disabled otherwise
1154 proxy_arp_pvlan - BOOLEAN
1155 Private VLAN proxy arp.
1156 Basically allow proxy arp replies back to the same interface
1157 (from which the ARP request/solicitation was received).
1159 This is done to support (ethernet) switch features, like RFC
1160 3069, where the individual ports are NOT allowed to
1161 communicate with each other, but they are allowed to talk to
1162 the upstream router. As described in RFC 3069, it is possible
1163 to allow these hosts to communicate through the upstream
1164 router by proxy_arp'ing. Don't need to be used together with
1167 This technology is known by different names:
1168 In RFC 3069 it is called VLAN Aggregation.
1169 Cisco and Allied Telesyn call it Private VLAN.
1170 Hewlett-Packard call it Source-Port filtering or port-isolation.
1171 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1173 shared_media - BOOLEAN
1174 Send(router) or accept(host) RFC1620 shared media redirects.
1175 Overrides secure_redirects.
1176 shared_media for the interface will be enabled if at least one of
1177 conf/{all,interface}/shared_media is set to TRUE,
1178 it will be disabled otherwise
1181 secure_redirects - BOOLEAN
1182 Accept ICMP redirect messages only to gateways listed in the
1183 interface's current gateway list. Even if disabled, RFC1122 redirect
1185 Overridden by shared_media.
1186 secure_redirects for the interface will be enabled if at least one of
1187 conf/{all,interface}/secure_redirects is set to TRUE,
1188 it will be disabled otherwise
1191 send_redirects - BOOLEAN
1192 Send redirects, if router.
1193 send_redirects for the interface will be enabled if at least one of
1194 conf/{all,interface}/send_redirects is set to TRUE,
1195 it will be disabled otherwise
1198 bootp_relay - BOOLEAN
1199 Accept packets with source address 0.b.c.d destined
1200 not to this host as local ones. It is supposed, that
1201 BOOTP relay daemon will catch and forward such packets.
1202 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1205 Not Implemented Yet.
1207 accept_source_route - BOOLEAN
1208 Accept packets with SRR option.
1209 conf/all/accept_source_route must also be set to TRUE to accept packets
1210 with SRR option on the interface
1211 default TRUE (router)
1214 accept_local - BOOLEAN
1215 Accept packets with local source addresses. In combination with
1216 suitable routing, this can be used to direct packets between two
1217 local interfaces over the wire and have them accepted properly.
1220 route_localnet - BOOLEAN
1221 Do not consider loopback addresses as martian source or destination
1222 while routing. This enables the use of 127/8 for local routing purposes.
1226 0 - No source validation.
1227 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1228 Each incoming packet is tested against the FIB and if the interface
1229 is not the best reverse path the packet check will fail.
1230 By default failed packets are discarded.
1231 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1232 Each incoming packet's source address is also tested against the FIB
1233 and if the source address is not reachable via any interface
1234 the packet check will fail.
1236 Current recommended practice in RFC3704 is to enable strict mode
1237 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1238 or other complicated routing, then loose mode is recommended.
1240 The max value from conf/{all,interface}/rp_filter is used
1241 when doing source validation on the {interface}.
1243 Default value is 0. Note that some distributions enable it
1246 arp_filter - BOOLEAN
1247 1 - Allows you to have multiple network interfaces on the same
1248 subnet, and have the ARPs for each interface be answered
1249 based on whether or not the kernel would route a packet from
1250 the ARP'd IP out that interface (therefore you must use source
1251 based routing for this to work). In other words it allows control
1252 of which cards (usually 1) will respond to an arp request.
1254 0 - (default) The kernel can respond to arp requests with addresses
1255 from other interfaces. This may seem wrong but it usually makes
1256 sense, because it increases the chance of successful communication.
1257 IP addresses are owned by the complete host on Linux, not by
1258 particular interfaces. Only for more complex setups like load-
1259 balancing, does this behaviour cause problems.
1261 arp_filter for the interface will be enabled if at least one of
1262 conf/{all,interface}/arp_filter is set to TRUE,
1263 it will be disabled otherwise
1265 arp_announce - INTEGER
1266 Define different restriction levels for announcing the local
1267 source IP address from IP packets in ARP requests sent on
1269 0 - (default) Use any local address, configured on any interface
1270 1 - Try to avoid local addresses that are not in the target's
1271 subnet for this interface. This mode is useful when target
1272 hosts reachable via this interface require the source IP
1273 address in ARP requests to be part of their logical network
1274 configured on the receiving interface. When we generate the
1275 request we will check all our subnets that include the
1276 target IP and will preserve the source address if it is from
1277 such subnet. If there is no such subnet we select source
1278 address according to the rules for level 2.
1279 2 - Always use the best local address for this target.
1280 In this mode we ignore the source address in the IP packet
1281 and try to select local address that we prefer for talks with
1282 the target host. Such local address is selected by looking
1283 for primary IP addresses on all our subnets on the outgoing
1284 interface that include the target IP address. If no suitable
1285 local address is found we select the first local address
1286 we have on the outgoing interface or on all other interfaces,
1287 with the hope we will receive reply for our request and
1288 even sometimes no matter the source IP address we announce.
1290 The max value from conf/{all,interface}/arp_announce is used.
1292 Increasing the restriction level gives more chance for
1293 receiving answer from the resolved target while decreasing
1294 the level announces more valid sender's information.
1296 arp_ignore - INTEGER
1297 Define different modes for sending replies in response to
1298 received ARP requests that resolve local target IP addresses:
1299 0 - (default): reply for any local target IP address, configured
1301 1 - reply only if the target IP address is local address
1302 configured on the incoming interface
1303 2 - reply only if the target IP address is local address
1304 configured on the incoming interface and both with the
1305 sender's IP address are part from same subnet on this interface
1306 3 - do not reply for local addresses configured with scope host,
1307 only resolutions for global and link addresses are replied
1309 8 - do not reply for all local addresses
1311 The max value from conf/{all,interface}/arp_ignore is used
1312 when ARP request is received on the {interface}
1314 arp_notify - BOOLEAN
1315 Define mode for notification of address and device changes.
1316 0 - (default): do nothing
1317 1 - Generate gratuitous arp requests when device is brought up
1318 or hardware address changes.
1320 arp_accept - BOOLEAN
1321 Define behavior for gratuitous ARP frames who's IP is not
1322 already present in the ARP table:
1323 0 - don't create new entries in the ARP table
1324 1 - create new entries in the ARP table
1326 Both replies and requests type gratuitous arp will trigger the
1327 ARP table to be updated, if this setting is on.
1329 If the ARP table already contains the IP address of the
1330 gratuitous arp frame, the arp table will be updated regardless
1331 if this setting is on or off.
1333 mcast_solicit - INTEGER
1334 The maximum number of multicast probes in INCOMPLETE state,
1335 when the associated hardware address is unknown. Defaults
1338 ucast_solicit - INTEGER
1339 The maximum number of unicast probes in PROBE state, when
1340 the hardware address is being reconfirmed. Defaults to 3.
1342 app_solicit - INTEGER
1343 The maximum number of probes to send to the user space ARP daemon
1344 via netlink before dropping back to multicast probes (see
1345 mcast_resolicit). Defaults to 0.
1347 mcast_resolicit - INTEGER
1348 The maximum number of multicast probes after unicast and
1349 app probes in PROBE state. Defaults to 0.
1351 disable_policy - BOOLEAN
1352 Disable IPSEC policy (SPD) for this interface
1354 disable_xfrm - BOOLEAN
1355 Disable IPSEC encryption on this interface, whatever the policy
1357 igmpv2_unsolicited_report_interval - INTEGER
1358 The interval in milliseconds in which the next unsolicited
1359 IGMPv1 or IGMPv2 report retransmit will take place.
1360 Default: 10000 (10 seconds)
1362 igmpv3_unsolicited_report_interval - INTEGER
1363 The interval in milliseconds in which the next unsolicited
1364 IGMPv3 report retransmit will take place.
1365 Default: 1000 (1 seconds)
1367 promote_secondaries - BOOLEAN
1368 When a primary IP address is removed from this interface
1369 promote a corresponding secondary IP address instead of
1370 removing all the corresponding secondary IP addresses.
1372 drop_unicast_in_l2_multicast - BOOLEAN
1373 Drop any unicast IP packets that are received in link-layer
1374 multicast (or broadcast) frames.
1375 This behavior (for multicast) is actually a SHOULD in RFC
1376 1122, but is disabled by default for compatibility reasons.
1379 drop_gratuitous_arp - BOOLEAN
1380 Drop all gratuitous ARP frames, for example if there's a known
1381 good ARP proxy on the network and such frames need not be used
1382 (or in the case of 802.11, must not be used to prevent attacks.)
1387 Allows you to write a number, which can be used as required.
1390 xfrm4_gc_thresh - INTEGER
1391 (Obsolete since linux-4.14)
1392 The threshold at which we will start garbage collecting for IPv4
1393 destination cache entries. At twice this value the system will
1394 refuse new allocations.
1396 igmp_link_local_mcast_reports - BOOLEAN
1397 Enable IGMP reports for link local multicast groups in the
1402 kuznet@ms2.inr.ac.ru
1408 delon.nicolas@wanadoo.fr
1413 /proc/sys/net/ipv6/* Variables:
1415 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1416 apply to IPv6 [XXX?].
1418 bindv6only - BOOLEAN
1419 Default value for IPV6_V6ONLY socket option,
1420 which restricts use of the IPv6 socket to IPv6 communication
1422 TRUE: disable IPv4-mapped address feature
1423 FALSE: enable IPv4-mapped address feature
1425 Default: FALSE (as specified in RFC3493)
1427 flowlabel_consistency - BOOLEAN
1428 Protect the consistency (and unicity) of flow label.
1429 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1435 auto_flowlabels - INTEGER
1436 Automatically generate flow labels based on a flow hash of the
1437 packet. This allows intermediate devices, such as routers, to
1438 identify packet flows for mechanisms like Equal Cost Multipath
1439 Routing (see RFC 6438).
1440 0: automatic flow labels are completely disabled
1441 1: automatic flow labels are enabled by default, they can be
1442 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1444 2: automatic flow labels are allowed, they may be enabled on a
1445 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1446 3: automatic flow labels are enabled and enforced, they cannot
1447 be disabled by the socket option
1450 flowlabel_state_ranges - BOOLEAN
1451 Split the flow label number space into two ranges. 0-0x7FFFF is
1452 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1453 is reserved for stateless flow labels as described in RFC6437.
1458 flowlabel_reflect - INTEGER
1459 Control flow label reflection. Needed for Path MTU
1460 Discovery to work with Equal Cost Multipath Routing in anycast
1461 environments. See RFC 7690 and:
1462 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1465 1: enabled for established flows
1467 Note that this prevents automatic flowlabel changes, as done
1468 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
1469 and "tcp: Change txhash on every SYN and RTO retransmit"
1471 2: enabled for TCP RESET packets (no active listener)
1472 If set, a RST packet sent in response to a SYN packet on a closed
1473 port will reflect the incoming flow label.
1475 4: enabled for ICMPv6 echo reply messages.
1479 fib_multipath_hash_policy - INTEGER
1480 Controls which hash policy to use for multipath routes.
1481 Default: 0 (Layer 3)
1483 0 - Layer 3 (source and destination addresses plus flow label)
1484 1 - Layer 4 (standard 5-tuple)
1485 2 - Layer 3 or inner Layer 3 if present
1487 anycast_src_echo_reply - BOOLEAN
1488 Controls the use of anycast addresses as source addresses for ICMPv6
1494 idgen_delay - INTEGER
1495 Controls the delay in seconds after which time to retry
1496 privacy stable address generation if a DAD conflict is
1498 Default: 1 (as specified in RFC7217)
1500 idgen_retries - INTEGER
1501 Controls the number of retries to generate a stable privacy
1502 address if a DAD conflict is detected.
1503 Default: 3 (as specified in RFC7217)
1506 Controls the MLD query robustness variable (see RFC3810 9.1).
1507 Default: 2 (as specified by RFC3810 9.1)
1508 Minimum: 1 (as specified by RFC6636 4.5)
1510 max_dst_opts_number - INTEGER
1511 Maximum number of non-padding TLVs allowed in a Destination
1512 options extension header. If this value is less than zero
1513 then unknown options are disallowed and the number of known
1514 TLVs allowed is the absolute value of this number.
1517 max_hbh_opts_number - INTEGER
1518 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1519 options extension header. If this value is less than zero
1520 then unknown options are disallowed and the number of known
1521 TLVs allowed is the absolute value of this number.
1524 max_dst_opts_length - INTEGER
1525 Maximum length allowed for a Destination options extension
1527 Default: INT_MAX (unlimited)
1529 max_hbh_length - INTEGER
1530 Maximum length allowed for a Hop-by-Hop options extension
1532 Default: INT_MAX (unlimited)
1534 skip_notify_on_dev_down - BOOLEAN
1535 Controls whether an RTM_DELROUTE message is generated for routes
1536 removed when a device is taken down or deleted. IPv4 does not
1537 generate this message; IPv6 does by default. Setting this sysctl
1538 to true skips the message, making IPv4 and IPv6 on par in relying
1539 on userspace caches to track link events and evict routes.
1540 Default: false (generate message)
1544 ip6frag_high_thresh - INTEGER
1545 Maximum memory used to reassemble IPv6 fragments. When
1546 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1547 the fragment handler will toss packets until ip6frag_low_thresh
1550 ip6frag_low_thresh - INTEGER
1551 See ip6frag_high_thresh
1553 ip6frag_time - INTEGER
1554 Time in seconds to keep an IPv6 fragment in memory.
1556 IPv6 Segment Routing:
1558 seg6_flowlabel - INTEGER
1559 Controls the behaviour of computing the flowlabel of outer
1560 IPv6 header in case of SR T.encaps
1562 -1 set flowlabel to zero.
1563 0 copy flowlabel from Inner packet in case of Inner IPv6
1564 (Set flowlabel to 0 in case IPv4/L2)
1565 1 Compute the flowlabel using seg6_make_flowlabel()
1570 Change the interface-specific default settings.
1574 Change all the interface-specific settings.
1576 [XXX: Other special features than forwarding?]
1578 conf/all/forwarding - BOOLEAN
1579 Enable global IPv6 forwarding between all interfaces.
1581 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1582 to control which interfaces may forward packets and which not.
1584 This also sets all interfaces' Host/Router setting
1585 'forwarding' to the specified value. See below for details.
1587 This referred to as global forwarding.
1592 fwmark_reflect - BOOLEAN
1593 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1594 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1595 If unset, these packets have a fwmark of zero. If set, they have the
1596 fwmark of the packet they are replying to.
1600 Change special settings per interface.
1602 The functional behaviour for certain settings is different
1603 depending on whether local forwarding is enabled or not.
1606 Accept Router Advertisements; autoconfigure using them.
1608 It also determines whether or not to transmit Router
1609 Solicitations. If and only if the functional setting is to
1610 accept Router Advertisements, Router Solicitations will be
1613 Possible values are:
1614 0 Do not accept Router Advertisements.
1615 1 Accept Router Advertisements if forwarding is disabled.
1616 2 Overrule forwarding behaviour. Accept Router Advertisements
1617 even if forwarding is enabled.
1619 Functional default: enabled if local forwarding is disabled.
1620 disabled if local forwarding is enabled.
1622 accept_ra_defrtr - BOOLEAN
1623 Learn default router in Router Advertisement.
1625 Functional default: enabled if accept_ra is enabled.
1626 disabled if accept_ra is disabled.
1628 accept_ra_from_local - BOOLEAN
1629 Accept RA with source-address that is found on local machine
1630 if the RA is otherwise proper and able to be accepted.
1631 Default is to NOT accept these as it may be an un-intended
1635 enabled if accept_ra_from_local is enabled
1636 on a specific interface.
1637 disabled if accept_ra_from_local is disabled
1638 on a specific interface.
1640 accept_ra_min_hop_limit - INTEGER
1641 Minimum hop limit Information in Router Advertisement.
1643 Hop limit Information in Router Advertisement less than this
1644 variable shall be ignored.
1648 accept_ra_pinfo - BOOLEAN
1649 Learn Prefix Information in Router Advertisement.
1651 Functional default: enabled if accept_ra is enabled.
1652 disabled if accept_ra is disabled.
1654 accept_ra_rt_info_min_plen - INTEGER
1655 Minimum prefix length of Route Information in RA.
1657 Route Information w/ prefix smaller than this variable shall
1660 Functional default: 0 if accept_ra_rtr_pref is enabled.
1661 -1 if accept_ra_rtr_pref is disabled.
1663 accept_ra_rt_info_max_plen - INTEGER
1664 Maximum prefix length of Route Information in RA.
1666 Route Information w/ prefix larger than this variable shall
1669 Functional default: 0 if accept_ra_rtr_pref is enabled.
1670 -1 if accept_ra_rtr_pref is disabled.
1672 accept_ra_rtr_pref - BOOLEAN
1673 Accept Router Preference in RA.
1675 Functional default: enabled if accept_ra is enabled.
1676 disabled if accept_ra is disabled.
1678 accept_ra_mtu - BOOLEAN
1679 Apply the MTU value specified in RA option 5 (RFC4861). If
1680 disabled, the MTU specified in the RA will be ignored.
1682 Functional default: enabled if accept_ra is enabled.
1683 disabled if accept_ra is disabled.
1685 accept_redirects - BOOLEAN
1688 Functional default: enabled if local forwarding is disabled.
1689 disabled if local forwarding is enabled.
1691 accept_source_route - INTEGER
1692 Accept source routing (routing extension header).
1694 >= 0: Accept only routing header type 2.
1695 < 0: Do not accept routing header.
1700 Autoconfigure addresses using Prefix Information in Router
1703 Functional default: enabled if accept_ra_pinfo is enabled.
1704 disabled if accept_ra_pinfo is disabled.
1706 dad_transmits - INTEGER
1707 The amount of Duplicate Address Detection probes to send.
1710 forwarding - INTEGER
1711 Configure interface-specific Host/Router behaviour.
1713 Note: It is recommended to have the same setting on all
1714 interfaces; mixed router/host scenarios are rather uncommon.
1716 Possible values are:
1717 0 Forwarding disabled
1718 1 Forwarding enabled
1722 By default, Host behaviour is assumed. This means:
1724 1. IsRouter flag is not set in Neighbour Advertisements.
1725 2. If accept_ra is TRUE (default), transmit Router
1727 3. If accept_ra is TRUE (default), accept Router
1728 Advertisements (and do autoconfiguration).
1729 4. If accept_redirects is TRUE (default), accept Redirects.
1733 If local forwarding is enabled, Router behaviour is assumed.
1734 This means exactly the reverse from the above:
1736 1. IsRouter flag is set in Neighbour Advertisements.
1737 2. Router Solicitations are not sent unless accept_ra is 2.
1738 3. Router Advertisements are ignored unless accept_ra is 2.
1739 4. Redirects are ignored.
1741 Default: 0 (disabled) if global forwarding is disabled (default),
1742 otherwise 1 (enabled).
1745 Default Hop Limit to set.
1749 Default Maximum Transfer Unit
1750 Default: 1280 (IPv6 required minimum)
1752 ip_nonlocal_bind - BOOLEAN
1753 If set, allows processes to bind() to non-local IPv6 addresses,
1754 which can be quite useful - but may break some applications.
1757 router_probe_interval - INTEGER
1758 Minimum interval (in seconds) between Router Probing described
1763 router_solicitation_delay - INTEGER
1764 Number of seconds to wait after interface is brought up
1765 before sending Router Solicitations.
1768 router_solicitation_interval - INTEGER
1769 Number of seconds to wait between Router Solicitations.
1772 router_solicitations - INTEGER
1773 Number of Router Solicitations to send until assuming no
1774 routers are present.
1777 use_oif_addrs_only - BOOLEAN
1778 When enabled, the candidate source addresses for destinations
1779 routed via this interface are restricted to the set of addresses
1780 configured on this interface (vis. RFC 6724, section 4).
1784 use_tempaddr - INTEGER
1785 Preference for Privacy Extensions (RFC3041).
1786 <= 0 : disable Privacy Extensions
1787 == 1 : enable Privacy Extensions, but prefer public
1788 addresses over temporary addresses.
1789 > 1 : enable Privacy Extensions and prefer temporary
1790 addresses over public addresses.
1791 Default: 0 (for most devices)
1792 -1 (for point-to-point devices and loopback devices)
1794 temp_valid_lft - INTEGER
1795 valid lifetime (in seconds) for temporary addresses.
1796 Default: 604800 (7 days)
1798 temp_prefered_lft - INTEGER
1799 Preferred lifetime (in seconds) for temporary addresses.
1800 Default: 86400 (1 day)
1802 keep_addr_on_down - INTEGER
1803 Keep all IPv6 addresses on an interface down event. If set static
1804 global addresses with no expiration time are not flushed.
1809 Default: 0 (addresses are removed)
1811 max_desync_factor - INTEGER
1812 Maximum value for DESYNC_FACTOR, which is a random value
1813 that ensures that clients don't synchronize with each
1814 other and generate new addresses at exactly the same time.
1815 value is in seconds.
1818 regen_max_retry - INTEGER
1819 Number of attempts before give up attempting to generate
1820 valid temporary addresses.
1823 max_addresses - INTEGER
1824 Maximum number of autoconfigured addresses per interface. Setting
1825 to zero disables the limitation. It is not recommended to set this
1826 value too large (or to zero) because it would be an easy way to
1827 crash the kernel by allowing too many addresses to be created.
1830 disable_ipv6 - BOOLEAN
1831 Disable IPv6 operation. If accept_dad is set to 2, this value
1832 will be dynamically set to TRUE if DAD fails for the link-local
1834 Default: FALSE (enable IPv6 operation)
1836 When this value is changed from 1 to 0 (IPv6 is being enabled),
1837 it will dynamically create a link-local address on the given
1838 interface and start Duplicate Address Detection, if necessary.
1840 When this value is changed from 0 to 1 (IPv6 is being disabled),
1841 it will dynamically delete all addresses and routes on the given
1842 interface. From now on it will not possible to add addresses/routes
1843 to the selected interface.
1845 accept_dad - INTEGER
1846 Whether to accept DAD (Duplicate Address Detection).
1848 1: Enable DAD (default)
1849 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1850 link-local address has been found.
1852 DAD operation and mode on a given interface will be selected according
1853 to the maximum value of conf/{all,interface}/accept_dad.
1855 force_tllao - BOOLEAN
1856 Enable sending the target link-layer address option even when
1857 responding to a unicast neighbor solicitation.
1860 Quoting from RFC 2461, section 4.4, Target link-layer address:
1862 "The option MUST be included for multicast solicitations in order to
1863 avoid infinite Neighbor Solicitation "recursion" when the peer node
1864 does not have a cache entry to return a Neighbor Advertisements
1865 message. When responding to unicast solicitations, the option can be
1866 omitted since the sender of the solicitation has the correct link-
1867 layer address; otherwise it would not have be able to send the unicast
1868 solicitation in the first place. However, including the link-layer
1869 address in this case adds little overhead and eliminates a potential
1870 race condition where the sender deletes the cached link-layer address
1871 prior to receiving a response to a previous solicitation."
1873 ndisc_notify - BOOLEAN
1874 Define mode for notification of address and device changes.
1875 0 - (default): do nothing
1876 1 - Generate unsolicited neighbour advertisements when device is brought
1877 up or hardware address changes.
1879 ndisc_tclass - INTEGER
1880 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
1881 Discovery (Router Solicitation, Router Advertisement, Neighbor
1882 Solicitation, Neighbor Advertisement, Redirect) messages.
1883 These 8 bits can be interpreted as 6 high order bits holding the DSCP
1884 value and 2 low order bits representing ECN (which you probably want
1888 mldv1_unsolicited_report_interval - INTEGER
1889 The interval in milliseconds in which the next unsolicited
1890 MLDv1 report retransmit will take place.
1891 Default: 10000 (10 seconds)
1893 mldv2_unsolicited_report_interval - INTEGER
1894 The interval in milliseconds in which the next unsolicited
1895 MLDv2 report retransmit will take place.
1896 Default: 1000 (1 second)
1898 force_mld_version - INTEGER
1899 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1900 1 - Enforce to use MLD version 1
1901 2 - Enforce to use MLD version 2
1903 suppress_frag_ndisc - INTEGER
1904 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1905 with IPv6 Neighbor Discovery) behavior:
1906 1 - (default) discard fragmented neighbor discovery packets
1907 0 - allow fragmented neighbor discovery packets
1909 optimistic_dad - BOOLEAN
1910 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1911 0: disabled (default)
1914 Optimistic Duplicate Address Detection for the interface will be enabled
1915 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
1916 it will be disabled otherwise.
1918 use_optimistic - BOOLEAN
1919 If enabled, do not classify optimistic addresses as deprecated during
1920 source address selection. Preferred addresses will still be chosen
1921 before optimistic addresses, subject to other ranking in the source
1922 address selection algorithm.
1923 0: disabled (default)
1926 This will be enabled if at least one of
1927 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
1929 stable_secret - IPv6 address
1930 This IPv6 address will be used as a secret to generate IPv6
1931 addresses for link-local addresses and autoconfigured
1932 ones. All addresses generated after setting this secret will
1933 be stable privacy ones by default. This can be changed via the
1934 addrgenmode ip-link. conf/default/stable_secret is used as the
1935 secret for the namespace, the interface specific ones can
1936 overwrite that. Writes to conf/all/stable_secret are refused.
1938 It is recommended to generate this secret during installation
1939 of a system and keep it stable after that.
1941 By default the stable secret is unset.
1943 addr_gen_mode - INTEGER
1944 Defines how link-local and autoconf addresses are generated.
1946 0: generate address based on EUI64 (default)
1947 1: do no generate a link-local address, use EUI64 for addresses generated
1949 2: generate stable privacy addresses, using the secret from
1950 stable_secret (RFC7217)
1951 3: generate stable privacy addresses, using a random secret if unset
1953 drop_unicast_in_l2_multicast - BOOLEAN
1954 Drop any unicast IPv6 packets that are received in link-layer
1955 multicast (or broadcast) frames.
1957 By default this is turned off.
1959 drop_unsolicited_na - BOOLEAN
1960 Drop all unsolicited neighbor advertisements, for example if there's
1961 a known good NA proxy on the network and such frames need not be used
1962 (or in the case of 802.11, must not be used to prevent attacks.)
1964 By default this is turned off.
1966 enhanced_dad - BOOLEAN
1967 Include a nonce option in the IPv6 neighbor solicitation messages used for
1968 duplicate address detection per RFC7527. A received DAD NS will only signal
1969 a duplicate address if the nonce is different. This avoids any false
1970 detection of duplicates due to loopback of the NS messages that we send.
1971 The nonce option will be sent on an interface unless both of
1972 conf/{all,interface}/enhanced_dad are set to FALSE.
1977 Limit the maximal rates for sending ICMPv6 messages.
1978 0 to disable any limiting,
1979 otherwise the minimal space between responses in milliseconds.
1982 ratemask - list of comma separated ranges
1983 For ICMPv6 message types matching the ranges in the ratemask, limit
1984 the sending of the message according to ratelimit parameter.
1986 The format used for both input and output is a comma separated
1987 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
1988 129). Writing to the file will clear all previous ranges of ICMPv6
1989 message types and update the current list with the input.
1991 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
1992 for numerical values of ICMPv6 message types, e.g. echo request is 128
1993 and echo reply is 129.
1995 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
1997 echo_ignore_all - BOOLEAN
1998 If set non-zero, then the kernel will ignore all ICMP ECHO
1999 requests sent to it over the IPv6 protocol.
2002 echo_ignore_multicast - BOOLEAN
2003 If set non-zero, then the kernel will ignore all ICMP ECHO
2004 requests sent to it over the IPv6 protocol via multicast.
2007 echo_ignore_anycast - BOOLEAN
2008 If set non-zero, then the kernel will ignore all ICMP ECHO
2009 requests sent to it over the IPv6 protocol destined to anycast address.
2012 xfrm6_gc_thresh - INTEGER
2013 (Obsolete since linux-4.14)
2014 The threshold at which we will start garbage collecting for IPv6
2015 destination cache entries. At twice this value the system will
2016 refuse new allocations.
2020 Pekka Savola <pekkas@netcore.fi>
2021 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2024 /proc/sys/net/bridge/* Variables:
2026 bridge-nf-call-arptables - BOOLEAN
2027 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2031 bridge-nf-call-iptables - BOOLEAN
2032 1 : pass bridged IPv4 traffic to iptables' chains.
2036 bridge-nf-call-ip6tables - BOOLEAN
2037 1 : pass bridged IPv6 traffic to ip6tables' chains.
2041 bridge-nf-filter-vlan-tagged - BOOLEAN
2042 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2046 bridge-nf-filter-pppoe-tagged - BOOLEAN
2047 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2051 bridge-nf-pass-vlan-input-dev - BOOLEAN
2052 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2053 interface on the bridge and set the netfilter input device to the vlan.
2054 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
2055 target work with vlan-on-top-of-bridge interfaces. When no matching
2056 vlan interface is found, or this switch is off, the input device is
2057 set to the bridge interface.
2058 0: disable bridge netfilter vlan interface lookup.
2061 proc/sys/net/sctp/* Variables:
2063 addip_enable - BOOLEAN
2064 Enable or disable extension of Dynamic Address Reconfiguration
2065 (ADD-IP) functionality specified in RFC5061. This extension provides
2066 the ability to dynamically add and remove new addresses for the SCTP
2069 1: Enable extension.
2071 0: Disable extension.
2076 Enable or disable pf (pf is short for potentially failed) state. A value
2077 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2078 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2079 Since pf_retrans and path_max_retrans can be changed by userspace
2080 application, sometimes user expects to disable pf state by the value of
2081 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2082 or path_max_retrans is changed by the user application, this pf state is
2083 enabled. As such, it is necessary to add this to dynamically enable
2084 and disable pf state. See:
2085 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2095 Unset or enable/disable pf (pf is short for potentially failed) state
2096 exposure. Applications can control the exposure of the PF path state
2097 in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2098 sockopt. When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2099 SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2100 can be got via SCTP_GET_PEER_ADDR_INFO sockopt; When it's enabled,
2101 a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2102 SCTP_PF state and a SCTP_PF-state transport info can be got via
2103 SCTP_GET_PEER_ADDR_INFO sockopt; When it's diabled, no
2104 SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2105 trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2108 0: Unset pf state exposure, Compatible with old applications.
2110 1: Disable pf state exposure.
2112 2: Enable pf state exposure.
2116 addip_noauth_enable - BOOLEAN
2117 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2118 authentication to protect the operations of adding or removing new
2119 addresses. This requirement is mandated so that unauthorized hosts
2120 would not be able to hijack associations. However, older
2121 implementations may not have implemented this requirement while
2122 allowing the ADD-IP extension. For reasons of interoperability,
2123 we provide this variable to control the enforcement of the
2124 authentication requirement.
2126 1: Allow ADD-IP extension to be used without authentication. This
2127 should only be set in a closed environment for interoperability
2128 with older implementations.
2130 0: Enforce the authentication requirement
2134 auth_enable - BOOLEAN
2135 Enable or disable Authenticated Chunks extension. This extension
2136 provides the ability to send and receive authenticated chunks and is
2137 required for secure operation of Dynamic Address Reconfiguration
2140 1: Enable this extension.
2141 0: Disable this extension.
2145 prsctp_enable - BOOLEAN
2146 Enable or disable the Partial Reliability extension (RFC3758) which
2147 is used to notify peers that a given DATA should no longer be expected.
2155 The limit of the number of new packets that can be initially sent. It
2156 controls how bursty the generated traffic can be.
2160 association_max_retrans - INTEGER
2161 Set the maximum number for retransmissions that an association can
2162 attempt deciding that the remote end is unreachable. If this value
2163 is exceeded, the association is terminated.
2167 max_init_retransmits - INTEGER
2168 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2169 that an association will attempt before declaring the destination
2170 unreachable and terminating.
2174 path_max_retrans - INTEGER
2175 The maximum number of retransmissions that will be attempted on a given
2176 path. Once this threshold is exceeded, the path is considered
2177 unreachable, and new traffic will use a different path when the
2178 association is multihomed.
2182 pf_retrans - INTEGER
2183 The number of retransmissions that will be attempted on a given path
2184 before traffic is redirected to an alternate transport (should one
2185 exist). Note this is distinct from path_max_retrans, as a path that
2186 passes the pf_retrans threshold can still be used. Its only
2187 deprioritized when a transmission path is selected by the stack. This
2188 setting is primarily used to enable fast failover mechanisms without
2189 having to reduce path_max_retrans to a very low value. See:
2190 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2191 for details. Note also that a value of pf_retrans > path_max_retrans
2192 disables this feature. Since both pf_retrans and path_max_retrans can
2193 be changed by userspace application, a variable pf_enable is used to
2198 ps_retrans - INTEGER
2199 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
2200 from section-5 "Primary Path Switchover" in rfc7829. The primary path
2201 will be changed to another active path when the path error counter on
2202 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
2203 to continue data transmission on a new working path even when the old
2204 primary destination address becomes active again". Note this feature
2205 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
2206 and its value can't be less than 'pf_retrans' when changing by sysctl.
2210 rto_initial - INTEGER
2211 The initial round trip timeout value in milliseconds that will be used
2212 in calculating round trip times. This is the initial time interval
2213 for retransmissions.
2218 The maximum value (in milliseconds) of the round trip timeout. This
2219 is the largest time interval that can elapse between retransmissions.
2224 The minimum value (in milliseconds) of the round trip timeout. This
2225 is the smallest time interval the can elapse between retransmissions.
2229 hb_interval - INTEGER
2230 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2231 are sent at the specified interval on idle paths to probe the state of
2232 a given path between 2 associations.
2236 sack_timeout - INTEGER
2237 The amount of time (in milliseconds) that the implementation will wait
2242 valid_cookie_life - INTEGER
2243 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2244 is used during association establishment.
2248 cookie_preserve_enable - BOOLEAN
2249 Enable or disable the ability to extend the lifetime of the SCTP cookie
2250 that is used during the establishment phase of SCTP association
2252 1: Enable cookie lifetime extension.
2257 cookie_hmac_alg - STRING
2258 Select the hmac algorithm used when generating the cookie value sent by
2259 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2264 Ability to assign md5 or sha1 as the selected alg is predicated on the
2265 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2266 CONFIG_CRYPTO_SHA1).
2268 Default: Dependent on configuration. MD5 if available, else SHA1 if
2269 available, else none.
2271 rcvbuf_policy - INTEGER
2272 Determines if the receive buffer is attributed to the socket or to
2273 association. SCTP supports the capability to create multiple
2274 associations on a single socket. When using this capability, it is
2275 possible that a single stalled association that's buffering a lot
2276 of data may block other associations from delivering their data by
2277 consuming all of the receive buffer space. To work around this,
2278 the rcvbuf_policy could be set to attribute the receiver buffer space
2279 to each association instead of the socket. This prevents the described
2282 1: rcvbuf space is per association
2283 0: rcvbuf space is per socket
2287 sndbuf_policy - INTEGER
2288 Similar to rcvbuf_policy above, this applies to send buffer space.
2290 1: Send buffer is tracked per association
2291 0: Send buffer is tracked per socket.
2295 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2296 Number of pages allowed for queueing by all SCTP sockets.
2298 min: Below this number of pages SCTP is not bothered about its
2299 memory appetite. When amount of memory allocated by SCTP exceeds
2300 this number, SCTP starts to moderate memory usage.
2302 pressure: This value was introduced to follow format of tcp_mem.
2304 max: Number of pages allowed for queueing by all SCTP sockets.
2306 Default is calculated at boot time from amount of available memory.
2308 sctp_rmem - vector of 3 INTEGERs: min, default, max
2309 Only the first value ("min") is used, "default" and "max" are
2312 min: Minimal size of receive buffer used by SCTP socket.
2313 It is guaranteed to each SCTP socket (but not association) even
2314 under moderate memory pressure.
2318 sctp_wmem - vector of 3 INTEGERs: min, default, max
2319 Currently this tunable has no effect.
2321 addr_scope_policy - INTEGER
2322 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2324 0 - Disable IPv4 address scoping
2325 1 - Enable IPv4 address scoping
2326 2 - Follow draft but allow IPv4 private addresses
2327 3 - Follow draft but allow IPv4 link local addresses
2332 /proc/sys/net/core/*
2333 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
2336 /proc/sys/net/unix/*
2337 max_dgram_qlen - INTEGER
2338 The maximum length of dgram socket receive queue