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 hardend 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 route/max_size - INTEGER
67 Maximum number of routes allowed in the kernel. Increase
68 this when using large numbers of interfaces and/or routes.
69 From linux kernel 3.6 onwards, this is deprecated for ipv4
70 as route cache is no longer used.
72 neigh/default/gc_thresh1 - INTEGER
73 Minimum number of entries to keep. Garbage collector will not
74 purge entries if there are fewer than this number.
77 neigh/default/gc_thresh2 - INTEGER
78 Threshold when garbage collector becomes more aggressive about
79 purging entries. Entries older than 5 seconds will be cleared
80 when over this number.
83 neigh/default/gc_thresh3 - INTEGER
84 Maximum number of neighbor entries allowed. Increase this
85 when using large numbers of interfaces and when communicating
86 with large numbers of directly-connected peers.
89 neigh/default/unres_qlen_bytes - INTEGER
90 The maximum number of bytes which may be used by packets
91 queued for each unresolved address by other network layers.
93 Setting negative value is meaningless and will return error.
94 Default: 65536 Bytes(64KB)
96 neigh/default/unres_qlen - INTEGER
97 The maximum number of packets which may be queued for each
98 unresolved address by other network layers.
99 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
100 Prior to linux 3.3, the default value is 3 which may cause
101 unexpected packet loss. The current default value is calculated
102 according to default value of unres_qlen_bytes and true size of
106 mtu_expires - INTEGER
107 Time, in seconds, that cached PMTU information is kept.
109 min_adv_mss - INTEGER
110 The advertised MSS depends on the first hop route MTU, but will
111 never be lower than this setting.
115 ipfrag_high_thresh - LONG INTEGER
116 Maximum memory used to reassemble IP fragments.
118 ipfrag_low_thresh - LONG INTEGER
119 (Obsolete since linux-4.4.174, backported from linux-4.17)
120 Maximum memory used to reassemble IP fragments before the kernel
121 begins to remove incomplete fragment queues to free up resources.
122 The kernel still accepts new fragments for defragmentation.
124 ipfrag_time - INTEGER
125 Time in seconds to keep an IP fragment in memory.
127 ipfrag_max_dist - INTEGER
128 ipfrag_max_dist is a non-negative integer value which defines the
129 maximum "disorder" which is allowed among fragments which share a
130 common IP source address. Note that reordering of packets is
131 not unusual, but if a large number of fragments arrive from a source
132 IP address while a particular fragment queue remains incomplete, it
133 probably indicates that one or more fragments belonging to that queue
134 have been lost. When ipfrag_max_dist is positive, an additional check
135 is done on fragments before they are added to a reassembly queue - if
136 ipfrag_max_dist (or more) fragments have arrived from a particular IP
137 address between additions to any IP fragment queue using that source
138 address, it's presumed that one or more fragments in the queue are
139 lost. The existing fragment queue will be dropped, and a new one
140 started. An ipfrag_max_dist value of zero disables this check.
142 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
143 result in unnecessarily dropping fragment queues when normal
144 reordering of packets occurs, which could lead to poor application
145 performance. Using a very large value, e.g. 50000, increases the
146 likelihood of incorrectly reassembling IP fragments that originate
147 from different IP datagrams, which could result in data corruption.
152 inet_peer_threshold - INTEGER
153 The approximate size of the storage. Starting from this threshold
154 entries will be thrown aggressively. This threshold also determines
155 entries' time-to-live and time intervals between garbage collection
156 passes. More entries, less time-to-live, less GC interval.
158 inet_peer_minttl - INTEGER
159 Minimum time-to-live of entries. Should be enough to cover fragment
160 time-to-live on the reassembling side. This minimum time-to-live is
161 guaranteed if the pool size is less than inet_peer_threshold.
164 inet_peer_maxttl - INTEGER
165 Maximum time-to-live of entries. Unused entries will expire after
166 this period of time if there is no memory pressure on the pool (i.e.
167 when the number of entries in the pool is very small).
173 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
174 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
177 tcp_abort_on_overflow - BOOLEAN
178 If listening service is too slow to accept new connections,
179 reset them. Default state is FALSE. It means that if overflow
180 occurred due to a burst, connection will recover. Enable this
181 option _only_ if you are really sure that listening daemon
182 cannot be tuned to accept connections faster. Enabling this
183 option can harm clients of your server.
185 tcp_adv_win_scale - INTEGER
186 Count buffering overhead as bytes/2^tcp_adv_win_scale
187 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
189 Possible values are [-31, 31], inclusive.
192 tcp_allowed_congestion_control - STRING
193 Show/set the congestion control choices available to non-privileged
194 processes. The list is a subset of those listed in
195 tcp_available_congestion_control.
196 Default is "reno" and the default setting (tcp_congestion_control).
198 tcp_app_win - INTEGER
199 Reserve max(window/2^tcp_app_win, mss) of window for application
200 buffer. Value 0 is special, it means that nothing is reserved.
203 tcp_autocorking - BOOLEAN
204 Enable TCP auto corking :
205 When applications do consecutive small write()/sendmsg() system calls,
206 we try to coalesce these small writes as much as possible, to lower
207 total amount of sent packets. This is done if at least one prior
208 packet for the flow is waiting in Qdisc queues or device transmit
209 queue. Applications can still use TCP_CORK for optimal behavior
210 when they know how/when to uncork their sockets.
213 tcp_available_congestion_control - STRING
214 Shows the available congestion control choices that are registered.
215 More congestion control algorithms may be available as modules,
218 tcp_base_mss - INTEGER
219 The initial value of search_low to be used by the packetization layer
220 Path MTU discovery (MTU probing). If MTU probing is enabled,
221 this is the initial MSS used by the connection.
223 tcp_min_snd_mss - INTEGER
224 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
225 as described in RFC 1122 and RFC 6691.
226 If this ADVMSS option is smaller than tcp_min_snd_mss,
227 it is silently capped to tcp_min_snd_mss.
229 Default : 48 (at least 8 bytes of payload per segment)
231 tcp_congestion_control - STRING
232 Set the congestion control algorithm to be used for new
233 connections. The algorithm "reno" is always available, but
234 additional choices may be available based on kernel configuration.
235 Default is set as part of kernel configuration.
236 For passive connections, the listener congestion control choice
238 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
241 Allows TCP to send "duplicate" SACKs.
243 tcp_early_retrans - INTEGER
244 Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
245 for triggering fast retransmit when the amount of outstanding data is
246 small and when no previously unsent data can be transmitted (such
247 that limited transmit could be used). Also controls the use of
248 Tail loss probe (TLP) that converts RTOs occurring due to tail
249 losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
253 2 enables ER but delays fast recovery and fast retransmit
254 by a fourth of RTT. This mitigates connection falsely
255 recovers when network has a small degree of reordering
256 (less than 3 packets).
257 3 enables delayed ER and TLP.
262 Control use of Explicit Congestion Notification (ECN) by TCP.
263 ECN is used only when both ends of the TCP connection indicate
264 support for it. This feature is useful in avoiding losses due
265 to congestion by allowing supporting routers to signal
266 congestion before having to drop packets.
268 0 Disable ECN. Neither initiate nor accept ECN.
269 1 Enable ECN when requested by incoming connections and
270 also request ECN on outgoing connection attempts.
271 2 Enable ECN when requested by incoming connections
272 but do not request ECN on outgoing connections.
275 tcp_ecn_fallback - BOOLEAN
276 If the kernel detects that ECN connection misbehaves, enable fall
277 back to non-ECN. Currently, this knob implements the fallback
278 from RFC3168, section 6.1.1.1., but we reserve that in future,
279 additional detection mechanisms could be implemented under this
280 knob. The value is not used, if tcp_ecn or per route (or congestion
281 control) ECN settings are disabled.
282 Default: 1 (fallback enabled)
285 Enable FACK congestion avoidance and fast retransmission.
286 The value is not used, if tcp_sack is not enabled.
288 tcp_fin_timeout - INTEGER
289 The length of time an orphaned (no longer referenced by any
290 application) connection will remain in the FIN_WAIT_2 state
291 before it is aborted at the local end. While a perfectly
292 valid "receive only" state for an un-orphaned connection, an
293 orphaned connection in FIN_WAIT_2 state could otherwise wait
294 forever for the remote to close its end of the connection.
299 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
300 F-RTO is an enhanced recovery algorithm for TCP retransmission
301 timeouts. It is particularly beneficial in networks where the
302 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
303 modification. It does not require any support from the peer.
305 By default it's enabled with a non-zero value. 0 disables F-RTO.
307 tcp_invalid_ratelimit - INTEGER
308 Limit the maximal rate for sending duplicate acknowledgments
309 in response to incoming TCP packets that are for an existing
310 connection but that are invalid due to any of these reasons:
312 (a) out-of-window sequence number,
313 (b) out-of-window acknowledgment number, or
314 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
316 This can help mitigate simple "ack loop" DoS attacks, wherein
317 a buggy or malicious middlebox or man-in-the-middle can
318 rewrite TCP header fields in manner that causes each endpoint
319 to think that the other is sending invalid TCP segments, thus
320 causing each side to send an unterminating stream of duplicate
321 acknowledgments for invalid segments.
323 Using 0 disables rate-limiting of dupacks in response to
324 invalid segments; otherwise this value specifies the minimal
325 space between sending such dupacks, in milliseconds.
327 Default: 500 (milliseconds).
329 tcp_keepalive_time - INTEGER
330 How often TCP sends out keepalive messages when keepalive is enabled.
333 tcp_keepalive_probes - INTEGER
334 How many keepalive probes TCP sends out, until it decides that the
335 connection is broken. Default value: 9.
337 tcp_keepalive_intvl - INTEGER
338 How frequently the probes are send out. Multiplied by
339 tcp_keepalive_probes it is time to kill not responding connection,
340 after probes started. Default value: 75sec i.e. connection
341 will be aborted after ~11 minutes of retries.
343 tcp_low_latency - BOOLEAN
344 If set, the TCP stack makes decisions that prefer lower
345 latency as opposed to higher throughput. By default, this
346 option is not set meaning that higher throughput is preferred.
347 An example of an application where this default should be
348 changed would be a Beowulf compute cluster.
351 tcp_max_orphans - INTEGER
352 Maximal number of TCP sockets not attached to any user file handle,
353 held by system. If this number is exceeded orphaned connections are
354 reset immediately and warning is printed. This limit exists
355 only to prevent simple DoS attacks, you _must_ not rely on this
356 or lower the limit artificially, but rather increase it
357 (probably, after increasing installed memory),
358 if network conditions require more than default value,
359 and tune network services to linger and kill such states
360 more aggressively. Let me to remind again: each orphan eats
361 up to ~64K of unswappable memory.
363 tcp_max_syn_backlog - INTEGER
364 Maximal number of remembered connection requests, which have not
365 received an acknowledgment from connecting client.
366 The minimal value is 128 for low memory machines, and it will
367 increase in proportion to the memory of machine.
368 If server suffers from overload, try increasing this number.
370 tcp_max_tw_buckets - INTEGER
371 Maximal number of timewait sockets held by system simultaneously.
372 If this number is exceeded time-wait socket is immediately destroyed
373 and warning is printed. This limit exists only to prevent
374 simple DoS attacks, you _must_ not lower the limit artificially,
375 but rather increase it (probably, after increasing installed memory),
376 if network conditions require more than default value.
378 tcp_mem - vector of 3 INTEGERs: min, pressure, max
379 min: below this number of pages TCP is not bothered about its
382 pressure: when amount of memory allocated by TCP exceeds this number
383 of pages, TCP moderates its memory consumption and enters memory
384 pressure mode, which is exited when memory consumption falls
387 max: number of pages allowed for queueing by all TCP sockets.
389 Defaults are calculated at boot time from amount of available
392 tcp_min_rtt_wlen - INTEGER
393 The window length of the windowed min filter to track the minimum RTT.
394 A shorter window lets a flow more quickly pick up new (higher)
395 minimum RTT when it is moved to a longer path (e.g., due to traffic
396 engineering). A longer window makes the filter more resistant to RTT
397 inflations such as transient congestion. The unit is seconds.
398 Possible values: 0 - 86400 (1 day)
401 tcp_moderate_rcvbuf - BOOLEAN
402 If set, TCP performs receive buffer auto-tuning, attempting to
403 automatically size the buffer (no greater than tcp_rmem[2]) to
404 match the size required by the path for full throughput. Enabled by
407 tcp_mtu_probing - INTEGER
408 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
411 1 - Disabled by default, enabled when an ICMP black hole detected
412 2 - Always enabled, use initial MSS of tcp_base_mss.
414 tcp_probe_interval - INTEGER
415 Controls how often to start TCP Packetization-Layer Path MTU
416 Discovery reprobe. The default is reprobing every 10 minutes as
419 tcp_probe_threshold - INTEGER
420 Controls when TCP Packetization-Layer Path MTU Discovery probing
421 will stop in respect to the width of search range in bytes. Default
424 tcp_no_metrics_save - BOOLEAN
425 By default, TCP saves various connection metrics in the route cache
426 when the connection closes, so that connections established in the
427 near future can use these to set initial conditions. Usually, this
428 increases overall performance, but may sometimes cause performance
429 degradation. If set, TCP will not cache metrics on closing
432 tcp_orphan_retries - INTEGER
433 This value influences the timeout of a locally closed TCP connection,
434 when RTO retransmissions remain unacknowledged.
435 See tcp_retries2 for more details.
437 The default value is 8.
438 If your machine is a loaded WEB server,
439 you should think about lowering this value, such sockets
440 may consume significant resources. Cf. tcp_max_orphans.
442 tcp_recovery - INTEGER
443 This value is a bitmap to enable various experimental loss recovery
446 RACK: 0x1 enables the RACK loss detection for fast detection of lost
447 retransmissions and tail drops.
451 tcp_reordering - INTEGER
452 Initial reordering level of packets in a TCP stream.
453 TCP stack can then dynamically adjust flow reordering level
454 between this initial value and tcp_max_reordering
457 tcp_max_reordering - INTEGER
458 Maximal reordering level of packets in a TCP stream.
459 300 is a fairly conservative value, but you might increase it
460 if paths are using per packet load balancing (like bonding rr mode)
463 tcp_retrans_collapse - BOOLEAN
464 Bug-to-bug compatibility with some broken printers.
465 On retransmit try to send bigger packets to work around bugs in
468 tcp_retries1 - INTEGER
469 This value influences the time, after which TCP decides, that
470 something is wrong due to unacknowledged RTO retransmissions,
471 and reports this suspicion to the network layer.
472 See tcp_retries2 for more details.
474 RFC 1122 recommends at least 3 retransmissions, which is the
477 tcp_retries2 - INTEGER
478 This value influences the timeout of an alive TCP connection,
479 when RTO retransmissions remain unacknowledged.
480 Given a value of N, a hypothetical TCP connection following
481 exponential backoff with an initial RTO of TCP_RTO_MIN would
482 retransmit N times before killing the connection at the (N+1)th RTO.
484 The default value of 15 yields a hypothetical timeout of 924.6
485 seconds and is a lower bound for the effective timeout.
486 TCP will effectively time out at the first RTO which exceeds the
487 hypothetical timeout.
489 RFC 1122 recommends at least 100 seconds for the timeout,
490 which corresponds to a value of at least 8.
492 tcp_rfc1337 - BOOLEAN
493 If set, the TCP stack behaves conforming to RFC1337. If unset,
494 we are not conforming to RFC, but prevent TCP TIME_WAIT
498 tcp_rmem - vector of 3 INTEGERs: min, default, max
499 min: Minimal size of receive buffer used by TCP sockets.
500 It is guaranteed to each TCP socket, even under moderate memory
504 default: initial size of receive buffer used by TCP sockets.
505 This value overrides net.core.rmem_default used by other protocols.
506 Default: 87380 bytes. This value results in window of 65535 with
507 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
508 less for default tcp_app_win. See below about these variables.
510 max: maximal size of receive buffer allowed for automatically
511 selected receiver buffers for TCP socket. This value does not override
512 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
513 automatic tuning of that socket's receive buffer size, in which
514 case this value is ignored.
515 Default: between 87380B and 6MB, depending on RAM size.
518 Enable select acknowledgments (SACKS).
520 tcp_slow_start_after_idle - BOOLEAN
521 If set, provide RFC2861 behavior and time out the congestion
522 window after an idle period. An idle period is defined at
523 the current RTO. If unset, the congestion window will not
524 be timed out after an idle period.
528 Use the Host requirements interpretation of the TCP urgent pointer field.
529 Most hosts use the older BSD interpretation, so if you turn this on
530 Linux might not communicate correctly with them.
533 tcp_synack_retries - INTEGER
534 Number of times SYNACKs for a passive TCP connection attempt will
535 be retransmitted. Should not be higher than 255. Default value
536 is 5, which corresponds to 31seconds till the last retransmission
537 with the current initial RTO of 1second. With this the final timeout
538 for a passive TCP connection will happen after 63seconds.
540 tcp_syncookies - BOOLEAN
541 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
542 Send out syncookies when the syn backlog queue of a socket
543 overflows. This is to prevent against the common 'SYN flood attack'
546 Note, that syncookies is fallback facility.
547 It MUST NOT be used to help highly loaded servers to stand
548 against legal connection rate. If you see SYN flood warnings
549 in your logs, but investigation shows that they occur
550 because of overload with legal connections, you should tune
551 another parameters until this warning disappear.
552 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
554 syncookies seriously violate TCP protocol, do not allow
555 to use TCP extensions, can result in serious degradation
556 of some services (f.e. SMTP relaying), visible not by you,
557 but your clients and relays, contacting you. While you see
558 SYN flood warnings in logs not being really flooded, your server
559 is seriously misconfigured.
561 If you want to test which effects syncookies have to your
562 network connections you can set this knob to 2 to enable
563 unconditionally generation of syncookies.
565 tcp_fastopen - INTEGER
566 Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
567 in the opening SYN packet. To use this feature, the client application
568 must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
569 connect() to perform a TCP handshake automatically.
571 The values (bitmap) are
572 1: Enables sending data in the opening SYN on the client w/ MSG_FASTOPEN.
573 2: Enables TCP Fast Open on the server side, i.e., allowing data in
574 a SYN packet to be accepted and passed to the application before
575 3-way hand shake finishes.
576 4: Send data in the opening SYN regardless of cookie availability and
577 without a cookie option.
578 0x100: Accept SYN data w/o validating the cookie.
579 0x200: Accept data-in-SYN w/o any cookie option present.
580 0x400/0x800: Enable Fast Open on all listeners regardless of the
581 TCP_FASTOPEN socket option. The two different flags designate two
582 different ways of setting max_qlen without the TCP_FASTOPEN socket
587 Note that the client & server side Fast Open flags (1 and 2
588 respectively) must be also enabled before the rest of flags can take
591 See include/net/tcp.h and the code for more details.
593 tcp_syn_retries - INTEGER
594 Number of times initial SYNs for an active TCP connection attempt
595 will be retransmitted. Should not be higher than 255. Default value
596 is 6, which corresponds to 63seconds till the last retransmission
597 with the current initial RTO of 1second. With this the final timeout
598 for an active TCP connection attempt will happen after 127seconds.
600 tcp_timestamps - BOOLEAN
601 Enable timestamps as defined in RFC1323.
603 tcp_min_tso_segs - INTEGER
604 Minimal number of segments per TSO frame.
605 Since linux-3.12, TCP does an automatic sizing of TSO frames,
606 depending on flow rate, instead of filling 64Kbytes packets.
607 For specific usages, it's possible to force TCP to build big
608 TSO frames. Note that TCP stack might split too big TSO packets
609 if available window is too small.
612 tcp_pacing_ss_ratio - INTEGER
613 sk->sk_pacing_rate is set by TCP stack using a ratio applied
614 to current rate. (current_rate = cwnd * mss / srtt)
615 If TCP is in slow start, tcp_pacing_ss_ratio is applied
616 to let TCP probe for bigger speeds, assuming cwnd can be
617 doubled every other RTT.
620 tcp_pacing_ca_ratio - INTEGER
621 sk->sk_pacing_rate is set by TCP stack using a ratio applied
622 to current rate. (current_rate = cwnd * mss / srtt)
623 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
624 is applied to conservatively probe for bigger throughput.
627 tcp_tso_win_divisor - INTEGER
628 This allows control over what percentage of the congestion window
629 can be consumed by a single TSO frame.
630 The setting of this parameter is a choice between burstiness and
631 building larger TSO frames.
634 tcp_tw_recycle - BOOLEAN
635 Enable fast recycling TIME-WAIT sockets. Default value is 0.
636 It should not be changed without advice/request of technical
639 tcp_tw_reuse - BOOLEAN
640 Allow to reuse TIME-WAIT sockets for new connections when it is
641 safe from protocol viewpoint. Default value is 0.
642 It should not be changed without advice/request of technical
645 tcp_window_scaling - BOOLEAN
646 Enable window scaling as defined in RFC1323.
648 tcp_wmem - vector of 3 INTEGERs: min, default, max
649 min: Amount of memory reserved for send buffers for TCP sockets.
650 Each TCP socket has rights to use it due to fact of its birth.
653 default: initial size of send buffer used by TCP sockets. This
654 value overrides net.core.wmem_default used by other protocols.
655 It is usually lower than net.core.wmem_default.
658 max: Maximal amount of memory allowed for automatically tuned
659 send buffers for TCP sockets. This value does not override
660 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
661 automatic tuning of that socket's send buffer size, in which case
662 this value is ignored.
663 Default: between 64K and 4MB, depending on RAM size.
665 tcp_notsent_lowat - UNSIGNED INTEGER
666 A TCP socket can control the amount of unsent bytes in its write queue,
667 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
668 reports POLLOUT events if the amount of unsent bytes is below a per
669 socket value, and if the write queue is not full. sendmsg() will
670 also not add new buffers if the limit is hit.
672 This global variable controls the amount of unsent data for
673 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
674 to the global variable has immediate effect.
676 Default: UINT_MAX (0xFFFFFFFF)
678 tcp_workaround_signed_windows - BOOLEAN
679 If set, assume no receipt of a window scaling option means the
680 remote TCP is broken and treats the window as a signed quantity.
681 If unset, assume the remote TCP is not broken even if we do
682 not receive a window scaling option from them.
685 tcp_thin_linear_timeouts - BOOLEAN
686 Enable dynamic triggering of linear timeouts for thin streams.
687 If set, a check is performed upon retransmission by timeout to
688 determine if the stream is thin (less than 4 packets in flight).
689 As long as the stream is found to be thin, up to 6 linear
690 timeouts may be performed before exponential backoff mode is
691 initiated. This improves retransmission latency for
692 non-aggressive thin streams, often found to be time-dependent.
693 For more information on thin streams, see
694 Documentation/networking/tcp-thin.txt
697 tcp_thin_dupack - BOOLEAN
698 Enable dynamic triggering of retransmissions after one dupACK
699 for thin streams. If set, a check is performed upon reception
700 of a dupACK to determine if the stream is thin (less than 4
701 packets in flight). As long as the stream is found to be thin,
702 data is retransmitted on the first received dupACK. This
703 improves retransmission latency for non-aggressive thin
704 streams, often found to be time-dependent.
705 For more information on thin streams, see
706 Documentation/networking/tcp-thin.txt
709 tcp_limit_output_bytes - INTEGER
710 Controls TCP Small Queue limit per tcp socket.
711 TCP bulk sender tends to increase packets in flight until it
712 gets losses notifications. With SNDBUF autotuning, this can
713 result in a large amount of packets queued in qdisc/device
714 on the local machine, hurting latency of other flows, for
715 typical pfifo_fast qdiscs.
716 tcp_limit_output_bytes limits the number of bytes on qdisc
717 or device to reduce artificial RTT/cwnd and reduce bufferbloat.
720 tcp_challenge_ack_limit - INTEGER
721 Limits number of Challenge ACK sent per second, as recommended
722 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
727 udp_mem - vector of 3 INTEGERs: min, pressure, max
728 Number of pages allowed for queueing by all UDP sockets.
730 min: Below this number of pages UDP is not bothered about its
731 memory appetite. When amount of memory allocated by UDP exceeds
732 this number, UDP starts to moderate memory usage.
734 pressure: This value was introduced to follow format of tcp_mem.
736 max: Number of pages allowed for queueing by all UDP sockets.
738 Default is calculated at boot time from amount of available memory.
740 udp_rmem_min - INTEGER
741 Minimal size of receive buffer used by UDP sockets in moderation.
742 Each UDP socket is able to use the size for receiving data, even if
743 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
746 udp_wmem_min - INTEGER
747 Minimal size of send buffer used by UDP sockets in moderation.
748 Each UDP socket is able to use the size for sending data, even if
749 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
754 cipso_cache_enable - BOOLEAN
755 If set, enable additions to and lookups from the CIPSO label mapping
756 cache. If unset, additions are ignored and lookups always result in a
757 miss. However, regardless of the setting the cache is still
758 invalidated when required when means you can safely toggle this on and
759 off and the cache will always be "safe".
762 cipso_cache_bucket_size - INTEGER
763 The CIPSO label cache consists of a fixed size hash table with each
764 hash bucket containing a number of cache entries. This variable limits
765 the number of entries in each hash bucket; the larger the value the
766 more CIPSO label mappings that can be cached. When the number of
767 entries in a given hash bucket reaches this limit adding new entries
768 causes the oldest entry in the bucket to be removed to make room.
771 cipso_rbm_optfmt - BOOLEAN
772 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
773 the CIPSO draft specification (see Documentation/netlabel for details).
774 This means that when set the CIPSO tag will be padded with empty
775 categories in order to make the packet data 32-bit aligned.
778 cipso_rbm_structvalid - BOOLEAN
779 If set, do a very strict check of the CIPSO option when
780 ip_options_compile() is called. If unset, relax the checks done during
781 ip_options_compile(). Either way is "safe" as errors are caught else
782 where in the CIPSO processing code but setting this to 0 (False) should
783 result in less work (i.e. it should be faster) but could cause problems
784 with other implementations that require strict checking.
789 ip_local_port_range - 2 INTEGERS
790 Defines the local port range that is used by TCP and UDP to
791 choose the local port. The first number is the first, the
792 second the last local port number.
793 If possible, it is better these numbers have different parity.
794 (one even and one odd values)
795 The default values are 32768 and 60999 respectively.
797 ip_local_reserved_ports - list of comma separated ranges
798 Specify the ports which are reserved for known third-party
799 applications. These ports will not be used by automatic port
800 assignments (e.g. when calling connect() or bind() with port
801 number 0). Explicit port allocation behavior is unchanged.
803 The format used for both input and output is a comma separated
804 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
805 10). Writing to the file will clear all previously reserved
806 ports and update the current list with the one given in the
809 Note that ip_local_port_range and ip_local_reserved_ports
810 settings are independent and both are considered by the kernel
811 when determining which ports are available for automatic port
814 You can reserve ports which are not in the current
815 ip_local_port_range, e.g.:
817 $ cat /proc/sys/net/ipv4/ip_local_port_range
819 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
822 although this is redundant. However such a setting is useful
823 if later the port range is changed to a value that will
824 include the reserved ports.
828 ip_nonlocal_bind - BOOLEAN
829 If set, allows processes to bind() to non-local IP addresses,
830 which can be quite useful - but may break some applications.
834 If set non-zero, enables support for dynamic addresses.
835 If set to a non-zero value larger than 1, a kernel log
836 message will be printed when dynamic address rewriting
840 ip_early_demux - BOOLEAN
841 Optimize input packet processing down to one demux for
842 certain kinds of local sockets. Currently we only do this
843 for established TCP sockets.
845 It may add an additional cost for pure routing workloads that
846 reduces overall throughput, in such case you should disable it.
849 icmp_echo_ignore_all - BOOLEAN
850 If set non-zero, then the kernel will ignore all ICMP ECHO
854 icmp_echo_ignore_broadcasts - BOOLEAN
855 If set non-zero, then the kernel will ignore all ICMP ECHO and
856 TIMESTAMP requests sent to it via broadcast/multicast.
859 icmp_ratelimit - INTEGER
860 Limit the maximal rates for sending ICMP packets whose type matches
861 icmp_ratemask (see below) to specific targets.
862 0 to disable any limiting,
863 otherwise the minimal space between responses in milliseconds.
864 Note that another sysctl, icmp_msgs_per_sec limits the number
865 of ICMP packets sent on all targets.
868 icmp_msgs_per_sec - INTEGER
869 Limit maximal number of ICMP packets sent per second from this host.
870 Only messages whose type matches icmp_ratemask (see below) are
871 controlled by this limit.
874 icmp_msgs_burst - INTEGER
875 icmp_msgs_per_sec controls number of ICMP packets sent per second,
876 while icmp_msgs_burst controls the burst size of these packets.
879 icmp_ratemask - INTEGER
880 Mask made of ICMP types for which rates are being limited.
881 Significant bits: IHGFEDCBA9876543210
882 Default mask: 0000001100000011000 (6168)
884 Bit definitions (see include/linux/icmp.h):
886 3 Destination Unreachable *
891 C Parameter Problem *
896 H Address Mask Request
899 * These are rate limited by default (see default mask above)
901 icmp_ignore_bogus_error_responses - BOOLEAN
902 Some routers violate RFC1122 by sending bogus responses to broadcast
903 frames. Such violations are normally logged via a kernel warning.
904 If this is set to TRUE, the kernel will not give such warnings, which
905 will avoid log file clutter.
908 icmp_errors_use_inbound_ifaddr - BOOLEAN
910 If zero, icmp error messages are sent with the primary address of
911 the exiting interface.
913 If non-zero, the message will be sent with the primary address of
914 the interface that received the packet that caused the icmp error.
915 This is the behaviour network many administrators will expect from
916 a router. And it can make debugging complicated network layouts
919 Note that if no primary address exists for the interface selected,
920 then the primary address of the first non-loopback interface that
921 has one will be used regardless of this setting.
925 igmp_max_memberships - INTEGER
926 Change the maximum number of multicast groups we can subscribe to.
929 Theoretical maximum value is bounded by having to send a membership
930 report in a single datagram (i.e. the report can't span multiple
931 datagrams, or risk confusing the switch and leaving groups you don't
934 The number of supported groups 'M' is bounded by the number of group
935 report entries you can fit into a single datagram of 65535 bytes.
937 M = 65536-sizeof (ip header)/(sizeof(Group record))
939 Group records are variable length, with a minimum of 12 bytes.
940 So net.ipv4.igmp_max_memberships should not be set higher than:
942 (65536-24) / 12 = 5459
944 The value 5459 assumes no IP header options, so in practice
945 this number may be lower.
947 conf/interface/* changes special settings per interface (where
948 "interface" is the name of your network interface)
950 conf/all/* is special, changes the settings for all interfaces
953 Controls the IGMP query robustness variable (see RFC2236 8.1).
954 Default: 2 (as specified by RFC2236 8.1)
955 Minimum: 1 (as specified by RFC6636 4.5)
957 log_martians - BOOLEAN
958 Log packets with impossible addresses to kernel log.
959 log_martians for the interface will be enabled if at least one of
960 conf/{all,interface}/log_martians is set to TRUE,
961 it will be disabled otherwise
963 accept_redirects - BOOLEAN
964 Accept ICMP redirect messages.
965 accept_redirects for the interface will be enabled if:
966 - both conf/{all,interface}/accept_redirects are TRUE in the case
967 forwarding for the interface is enabled
969 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
970 case forwarding for the interface is disabled
971 accept_redirects for the interface will be disabled otherwise
976 Enable IP forwarding on this interface.
978 mc_forwarding - BOOLEAN
979 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
980 and a multicast routing daemon is required.
981 conf/all/mc_forwarding must also be set to TRUE to enable multicast
982 routing for the interface
985 Integer value used to differentiate the devices by the medium they
986 are attached to. Two devices can have different id values when
987 the broadcast packets are received only on one of them.
988 The default value 0 means that the device is the only interface
989 to its medium, value of -1 means that medium is not known.
991 Currently, it is used to change the proxy_arp behavior:
992 the proxy_arp feature is enabled for packets forwarded between
993 two devices attached to different media.
997 proxy_arp for the interface will be enabled if at least one of
998 conf/{all,interface}/proxy_arp is set to TRUE,
999 it will be disabled otherwise
1001 proxy_arp_pvlan - BOOLEAN
1002 Private VLAN proxy arp.
1003 Basically allow proxy arp replies back to the same interface
1004 (from which the ARP request/solicitation was received).
1006 This is done to support (ethernet) switch features, like RFC
1007 3069, where the individual ports are NOT allowed to
1008 communicate with each other, but they are allowed to talk to
1009 the upstream router. As described in RFC 3069, it is possible
1010 to allow these hosts to communicate through the upstream
1011 router by proxy_arp'ing. Don't need to be used together with
1014 This technology is known by different names:
1015 In RFC 3069 it is called VLAN Aggregation.
1016 Cisco and Allied Telesyn call it Private VLAN.
1017 Hewlett-Packard call it Source-Port filtering or port-isolation.
1018 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1020 shared_media - BOOLEAN
1021 Send(router) or accept(host) RFC1620 shared media redirects.
1022 Overrides ip_secure_redirects.
1023 shared_media for the interface will be enabled if at least one of
1024 conf/{all,interface}/shared_media is set to TRUE,
1025 it will be disabled otherwise
1028 secure_redirects - BOOLEAN
1029 Accept ICMP redirect messages only for gateways,
1030 listed in default gateway list.
1031 secure_redirects for the interface will be enabled if at least one of
1032 conf/{all,interface}/secure_redirects is set to TRUE,
1033 it will be disabled otherwise
1036 send_redirects - BOOLEAN
1037 Send redirects, if router.
1038 send_redirects for the interface will be enabled if at least one of
1039 conf/{all,interface}/send_redirects is set to TRUE,
1040 it will be disabled otherwise
1043 bootp_relay - BOOLEAN
1044 Accept packets with source address 0.b.c.d destined
1045 not to this host as local ones. It is supposed, that
1046 BOOTP relay daemon will catch and forward such packets.
1047 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1050 Not Implemented Yet.
1052 accept_source_route - BOOLEAN
1053 Accept packets with SRR option.
1054 conf/all/accept_source_route must also be set to TRUE to accept packets
1055 with SRR option on the interface
1056 default TRUE (router)
1059 accept_local - BOOLEAN
1060 Accept packets with local source addresses. In combination with
1061 suitable routing, this can be used to direct packets between two
1062 local interfaces over the wire and have them accepted properly.
1065 route_localnet - BOOLEAN
1066 Do not consider loopback addresses as martian source or destination
1067 while routing. This enables the use of 127/8 for local routing purposes.
1071 0 - No source validation.
1072 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1073 Each incoming packet is tested against the FIB and if the interface
1074 is not the best reverse path the packet check will fail.
1075 By default failed packets are discarded.
1076 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1077 Each incoming packet's source address is also tested against the FIB
1078 and if the source address is not reachable via any interface
1079 the packet check will fail.
1081 Current recommended practice in RFC3704 is to enable strict mode
1082 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1083 or other complicated routing, then loose mode is recommended.
1085 The max value from conf/{all,interface}/rp_filter is used
1086 when doing source validation on the {interface}.
1088 Default value is 0. Note that some distributions enable it
1091 arp_filter - BOOLEAN
1092 1 - Allows you to have multiple network interfaces on the same
1093 subnet, and have the ARPs for each interface be answered
1094 based on whether or not the kernel would route a packet from
1095 the ARP'd IP out that interface (therefore you must use source
1096 based routing for this to work). In other words it allows control
1097 of which cards (usually 1) will respond to an arp request.
1099 0 - (default) The kernel can respond to arp requests with addresses
1100 from other interfaces. This may seem wrong but it usually makes
1101 sense, because it increases the chance of successful communication.
1102 IP addresses are owned by the complete host on Linux, not by
1103 particular interfaces. Only for more complex setups like load-
1104 balancing, does this behaviour cause problems.
1106 arp_filter for the interface will be enabled if at least one of
1107 conf/{all,interface}/arp_filter is set to TRUE,
1108 it will be disabled otherwise
1110 arp_announce - INTEGER
1111 Define different restriction levels for announcing the local
1112 source IP address from IP packets in ARP requests sent on
1114 0 - (default) Use any local address, configured on any interface
1115 1 - Try to avoid local addresses that are not in the target's
1116 subnet for this interface. This mode is useful when target
1117 hosts reachable via this interface require the source IP
1118 address in ARP requests to be part of their logical network
1119 configured on the receiving interface. When we generate the
1120 request we will check all our subnets that include the
1121 target IP and will preserve the source address if it is from
1122 such subnet. If there is no such subnet we select source
1123 address according to the rules for level 2.
1124 2 - Always use the best local address for this target.
1125 In this mode we ignore the source address in the IP packet
1126 and try to select local address that we prefer for talks with
1127 the target host. Such local address is selected by looking
1128 for primary IP addresses on all our subnets on the outgoing
1129 interface that include the target IP address. If no suitable
1130 local address is found we select the first local address
1131 we have on the outgoing interface or on all other interfaces,
1132 with the hope we will receive reply for our request and
1133 even sometimes no matter the source IP address we announce.
1135 The max value from conf/{all,interface}/arp_announce is used.
1137 Increasing the restriction level gives more chance for
1138 receiving answer from the resolved target while decreasing
1139 the level announces more valid sender's information.
1141 arp_ignore - INTEGER
1142 Define different modes for sending replies in response to
1143 received ARP requests that resolve local target IP addresses:
1144 0 - (default): reply for any local target IP address, configured
1146 1 - reply only if the target IP address is local address
1147 configured on the incoming interface
1148 2 - reply only if the target IP address is local address
1149 configured on the incoming interface and both with the
1150 sender's IP address are part from same subnet on this interface
1151 3 - do not reply for local addresses configured with scope host,
1152 only resolutions for global and link addresses are replied
1154 8 - do not reply for all local addresses
1156 The max value from conf/{all,interface}/arp_ignore is used
1157 when ARP request is received on the {interface}
1159 arp_notify - BOOLEAN
1160 Define mode for notification of address and device changes.
1161 0 - (default): do nothing
1162 1 - Generate gratuitous arp requests when device is brought up
1163 or hardware address changes.
1165 arp_accept - BOOLEAN
1166 Define behavior for gratuitous ARP frames who's IP is not
1167 already present in the ARP table:
1168 0 - don't create new entries in the ARP table
1169 1 - create new entries in the ARP table
1171 Both replies and requests type gratuitous arp will trigger the
1172 ARP table to be updated, if this setting is on.
1174 If the ARP table already contains the IP address of the
1175 gratuitous arp frame, the arp table will be updated regardless
1176 if this setting is on or off.
1178 mcast_solicit - INTEGER
1179 The maximum number of multicast probes in INCOMPLETE state,
1180 when the associated hardware address is unknown. Defaults
1183 ucast_solicit - INTEGER
1184 The maximum number of unicast probes in PROBE state, when
1185 the hardware address is being reconfirmed. Defaults to 3.
1187 app_solicit - INTEGER
1188 The maximum number of probes to send to the user space ARP daemon
1189 via netlink before dropping back to multicast probes (see
1190 mcast_resolicit). Defaults to 0.
1192 mcast_resolicit - INTEGER
1193 The maximum number of multicast probes after unicast and
1194 app probes in PROBE state. Defaults to 0.
1196 disable_policy - BOOLEAN
1197 Disable IPSEC policy (SPD) for this interface
1199 disable_xfrm - BOOLEAN
1200 Disable IPSEC encryption on this interface, whatever the policy
1202 igmpv2_unsolicited_report_interval - INTEGER
1203 The interval in milliseconds in which the next unsolicited
1204 IGMPv1 or IGMPv2 report retransmit will take place.
1205 Default: 10000 (10 seconds)
1207 igmpv3_unsolicited_report_interval - INTEGER
1208 The interval in milliseconds in which the next unsolicited
1209 IGMPv3 report retransmit will take place.
1210 Default: 1000 (1 seconds)
1212 promote_secondaries - BOOLEAN
1213 When a primary IP address is removed from this interface
1214 promote a corresponding secondary IP address instead of
1215 removing all the corresponding secondary IP addresses.
1219 Allows you to write a number, which can be used as required.
1222 xfrm4_gc_thresh - INTEGER
1223 The threshold at which we will start garbage collecting for IPv4
1224 destination cache entries. At twice this value the system will
1225 refuse new allocations. The value must be set below the flowcache
1226 limit (4096 * number of online cpus) to take effect.
1228 igmp_link_local_mcast_reports - BOOLEAN
1229 Enable IGMP reports for link local multicast groups in the
1234 kuznet@ms2.inr.ac.ru
1240 delon.nicolas@wanadoo.fr
1245 /proc/sys/net/ipv6/* Variables:
1247 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1248 apply to IPv6 [XXX?].
1250 bindv6only - BOOLEAN
1251 Default value for IPV6_V6ONLY socket option,
1252 which restricts use of the IPv6 socket to IPv6 communication
1254 TRUE: disable IPv4-mapped address feature
1255 FALSE: enable IPv4-mapped address feature
1257 Default: FALSE (as specified in RFC3493)
1259 flowlabel_consistency - BOOLEAN
1260 Protect the consistency (and unicity) of flow label.
1261 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1267 auto_flowlabels - INTEGER
1268 Automatically generate flow labels based on a flow hash of the
1269 packet. This allows intermediate devices, such as routers, to
1270 identify packet flows for mechanisms like Equal Cost Multipath
1271 Routing (see RFC 6438).
1272 0: automatic flow labels are completely disabled
1273 1: automatic flow labels are enabled by default, they can be
1274 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1276 2: automatic flow labels are allowed, they may be enabled on a
1277 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1278 3: automatic flow labels are enabled and enforced, they cannot
1279 be disabled by the socket option
1282 flowlabel_state_ranges - BOOLEAN
1283 Split the flow label number space into two ranges. 0-0x7FFFF is
1284 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1285 is reserved for stateless flow labels as described in RFC6437.
1290 anycast_src_echo_reply - BOOLEAN
1291 Controls the use of anycast addresses as source addresses for ICMPv6
1297 idgen_delay - INTEGER
1298 Controls the delay in seconds after which time to retry
1299 privacy stable address generation if a DAD conflict is
1301 Default: 1 (as specified in RFC7217)
1303 idgen_retries - INTEGER
1304 Controls the number of retries to generate a stable privacy
1305 address if a DAD conflict is detected.
1306 Default: 3 (as specified in RFC7217)
1309 Controls the MLD query robustness variable (see RFC3810 9.1).
1310 Default: 2 (as specified by RFC3810 9.1)
1311 Minimum: 1 (as specified by RFC6636 4.5)
1315 ip6frag_high_thresh - INTEGER
1316 Maximum memory used to reassemble IPv6 fragments. When
1317 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1318 the fragment handler will toss packets until ip6frag_low_thresh
1321 ip6frag_low_thresh - INTEGER
1322 See ip6frag_high_thresh
1324 ip6frag_time - INTEGER
1325 Time in seconds to keep an IPv6 fragment in memory.
1328 Change the interface-specific default settings.
1332 Change all the interface-specific settings.
1334 [XXX: Other special features than forwarding?]
1336 conf/all/forwarding - BOOLEAN
1337 Enable global IPv6 forwarding between all interfaces.
1339 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1340 to control which interfaces may forward packets and which not.
1342 This also sets all interfaces' Host/Router setting
1343 'forwarding' to the specified value. See below for details.
1345 This referred to as global forwarding.
1350 fwmark_reflect - BOOLEAN
1351 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1352 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1353 If unset, these packets have a fwmark of zero. If set, they have the
1354 fwmark of the packet they are replying to.
1358 Change special settings per interface.
1360 The functional behaviour for certain settings is different
1361 depending on whether local forwarding is enabled or not.
1364 Accept Router Advertisements; autoconfigure using them.
1366 It also determines whether or not to transmit Router
1367 Solicitations. If and only if the functional setting is to
1368 accept Router Advertisements, Router Solicitations will be
1371 Possible values are:
1372 0 Do not accept Router Advertisements.
1373 1 Accept Router Advertisements if forwarding is disabled.
1374 2 Overrule forwarding behaviour. Accept Router Advertisements
1375 even if forwarding is enabled.
1377 Functional default: enabled if local forwarding is disabled.
1378 disabled if local forwarding is enabled.
1380 accept_ra_defrtr - BOOLEAN
1381 Learn default router in Router Advertisement.
1383 Functional default: enabled if accept_ra is enabled.
1384 disabled if accept_ra is disabled.
1386 accept_ra_from_local - BOOLEAN
1387 Accept RA with source-address that is found on local machine
1388 if the RA is otherwise proper and able to be accepted.
1389 Default is to NOT accept these as it may be an un-intended
1393 enabled if accept_ra_from_local is enabled
1394 on a specific interface.
1395 disabled if accept_ra_from_local is disabled
1396 on a specific interface.
1398 accept_ra_min_hop_limit - INTEGER
1399 Minimum hop limit Information in Router Advertisement.
1401 Hop limit Information in Router Advertisement less than this
1402 variable shall be ignored.
1406 accept_ra_pinfo - BOOLEAN
1407 Learn Prefix Information in Router Advertisement.
1409 Functional default: enabled if accept_ra is enabled.
1410 disabled if accept_ra is disabled.
1412 accept_ra_rt_info_max_plen - INTEGER
1413 Maximum prefix length of Route Information in RA.
1415 Route Information w/ prefix larger than or equal to this
1416 variable shall be ignored.
1418 Functional default: 0 if accept_ra_rtr_pref is enabled.
1419 -1 if accept_ra_rtr_pref is disabled.
1421 accept_ra_rtr_pref - BOOLEAN
1422 Accept Router Preference in RA.
1424 Functional default: enabled if accept_ra is enabled.
1425 disabled if accept_ra is disabled.
1427 accept_ra_mtu - BOOLEAN
1428 Apply the MTU value specified in RA option 5 (RFC4861). If
1429 disabled, the MTU specified in the RA will be ignored.
1431 Functional default: enabled if accept_ra is enabled.
1432 disabled if accept_ra is disabled.
1434 accept_redirects - BOOLEAN
1437 Functional default: enabled if local forwarding is disabled.
1438 disabled if local forwarding is enabled.
1440 accept_source_route - INTEGER
1441 Accept source routing (routing extension header).
1443 >= 0: Accept only routing header type 2.
1444 < 0: Do not accept routing header.
1449 Autoconfigure addresses using Prefix Information in Router
1452 Functional default: enabled if accept_ra_pinfo is enabled.
1453 disabled if accept_ra_pinfo is disabled.
1455 dad_transmits - INTEGER
1456 The amount of Duplicate Address Detection probes to send.
1459 forwarding - INTEGER
1460 Configure interface-specific Host/Router behaviour.
1462 Note: It is recommended to have the same setting on all
1463 interfaces; mixed router/host scenarios are rather uncommon.
1465 Possible values are:
1466 0 Forwarding disabled
1467 1 Forwarding enabled
1471 By default, Host behaviour is assumed. This means:
1473 1. IsRouter flag is not set in Neighbour Advertisements.
1474 2. If accept_ra is TRUE (default), transmit Router
1476 3. If accept_ra is TRUE (default), accept Router
1477 Advertisements (and do autoconfiguration).
1478 4. If accept_redirects is TRUE (default), accept Redirects.
1482 If local forwarding is enabled, Router behaviour is assumed.
1483 This means exactly the reverse from the above:
1485 1. IsRouter flag is set in Neighbour Advertisements.
1486 2. Router Solicitations are not sent unless accept_ra is 2.
1487 3. Router Advertisements are ignored unless accept_ra is 2.
1488 4. Redirects are ignored.
1490 Default: 0 (disabled) if global forwarding is disabled (default),
1491 otherwise 1 (enabled).
1494 Default Hop Limit to set.
1498 Default Maximum Transfer Unit
1499 Default: 1280 (IPv6 required minimum)
1501 ip_nonlocal_bind - BOOLEAN
1502 If set, allows processes to bind() to non-local IPv6 addresses,
1503 which can be quite useful - but may break some applications.
1506 router_probe_interval - INTEGER
1507 Minimum interval (in seconds) between Router Probing described
1512 router_solicitation_delay - INTEGER
1513 Number of seconds to wait after interface is brought up
1514 before sending Router Solicitations.
1517 router_solicitation_interval - INTEGER
1518 Number of seconds to wait between Router Solicitations.
1521 router_solicitations - INTEGER
1522 Number of Router Solicitations to send until assuming no
1523 routers are present.
1526 use_oif_addrs_only - BOOLEAN
1527 When enabled, the candidate source addresses for destinations
1528 routed via this interface are restricted to the set of addresses
1529 configured on this interface (vis. RFC 6724, section 4).
1533 use_tempaddr - INTEGER
1534 Preference for Privacy Extensions (RFC3041).
1535 <= 0 : disable Privacy Extensions
1536 == 1 : enable Privacy Extensions, but prefer public
1537 addresses over temporary addresses.
1538 > 1 : enable Privacy Extensions and prefer temporary
1539 addresses over public addresses.
1540 Default: 0 (for most devices)
1541 -1 (for point-to-point devices and loopback devices)
1543 temp_valid_lft - INTEGER
1544 valid lifetime (in seconds) for temporary addresses.
1545 Default: 604800 (7 days)
1547 temp_prefered_lft - INTEGER
1548 Preferred lifetime (in seconds) for temporary addresses.
1549 Default: 86400 (1 day)
1551 max_desync_factor - INTEGER
1552 Maximum value for DESYNC_FACTOR, which is a random value
1553 that ensures that clients don't synchronize with each
1554 other and generate new addresses at exactly the same time.
1555 value is in seconds.
1558 regen_max_retry - INTEGER
1559 Number of attempts before give up attempting to generate
1560 valid temporary addresses.
1563 max_addresses - INTEGER
1564 Maximum number of autoconfigured addresses per interface. Setting
1565 to zero disables the limitation. It is not recommended to set this
1566 value too large (or to zero) because it would be an easy way to
1567 crash the kernel by allowing too many addresses to be created.
1570 disable_ipv6 - BOOLEAN
1571 Disable IPv6 operation. If accept_dad is set to 2, this value
1572 will be dynamically set to TRUE if DAD fails for the link-local
1574 Default: FALSE (enable IPv6 operation)
1576 When this value is changed from 1 to 0 (IPv6 is being enabled),
1577 it will dynamically create a link-local address on the given
1578 interface and start Duplicate Address Detection, if necessary.
1580 When this value is changed from 0 to 1 (IPv6 is being disabled),
1581 it will dynamically delete all address on the given interface.
1583 accept_dad - INTEGER
1584 Whether to accept DAD (Duplicate Address Detection).
1586 1: Enable DAD (default)
1587 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1588 link-local address has been found.
1590 force_tllao - BOOLEAN
1591 Enable sending the target link-layer address option even when
1592 responding to a unicast neighbor solicitation.
1595 Quoting from RFC 2461, section 4.4, Target link-layer address:
1597 "The option MUST be included for multicast solicitations in order to
1598 avoid infinite Neighbor Solicitation "recursion" when the peer node
1599 does not have a cache entry to return a Neighbor Advertisements
1600 message. When responding to unicast solicitations, the option can be
1601 omitted since the sender of the solicitation has the correct link-
1602 layer address; otherwise it would not have be able to send the unicast
1603 solicitation in the first place. However, including the link-layer
1604 address in this case adds little overhead and eliminates a potential
1605 race condition where the sender deletes the cached link-layer address
1606 prior to receiving a response to a previous solicitation."
1608 ndisc_notify - BOOLEAN
1609 Define mode for notification of address and device changes.
1610 0 - (default): do nothing
1611 1 - Generate unsolicited neighbour advertisements when device is brought
1612 up or hardware address changes.
1614 mldv1_unsolicited_report_interval - INTEGER
1615 The interval in milliseconds in which the next unsolicited
1616 MLDv1 report retransmit will take place.
1617 Default: 10000 (10 seconds)
1619 mldv2_unsolicited_report_interval - INTEGER
1620 The interval in milliseconds in which the next unsolicited
1621 MLDv2 report retransmit will take place.
1622 Default: 1000 (1 second)
1624 force_mld_version - INTEGER
1625 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1626 1 - Enforce to use MLD version 1
1627 2 - Enforce to use MLD version 2
1629 suppress_frag_ndisc - INTEGER
1630 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1631 with IPv6 Neighbor Discovery) behavior:
1632 1 - (default) discard fragmented neighbor discovery packets
1633 0 - allow fragmented neighbor discovery packets
1635 optimistic_dad - BOOLEAN
1636 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1637 0: disabled (default)
1640 use_optimistic - BOOLEAN
1641 If enabled, do not classify optimistic addresses as deprecated during
1642 source address selection. Preferred addresses will still be chosen
1643 before optimistic addresses, subject to other ranking in the source
1644 address selection algorithm.
1645 0: disabled (default)
1648 stable_secret - IPv6 address
1649 This IPv6 address will be used as a secret to generate IPv6
1650 addresses for link-local addresses and autoconfigured
1651 ones. All addresses generated after setting this secret will
1652 be stable privacy ones by default. This can be changed via the
1653 addrgenmode ip-link. conf/default/stable_secret is used as the
1654 secret for the namespace, the interface specific ones can
1655 overwrite that. Writes to conf/all/stable_secret are refused.
1657 It is recommended to generate this secret during installation
1658 of a system and keep it stable after that.
1660 By default the stable secret is unset.
1664 Limit the maximal rates for sending ICMPv6 packets.
1665 0 to disable any limiting,
1666 otherwise the minimal space between responses in milliseconds.
1669 xfrm6_gc_thresh - INTEGER
1670 The threshold at which we will start garbage collecting for IPv6
1671 destination cache entries. At twice this value the system will
1672 refuse new allocations. The value must be set below the flowcache
1673 limit (4096 * number of online cpus) to take effect.
1677 Pekka Savola <pekkas@netcore.fi>
1678 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1681 /proc/sys/net/bridge/* Variables:
1683 bridge-nf-call-arptables - BOOLEAN
1684 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1688 bridge-nf-call-iptables - BOOLEAN
1689 1 : pass bridged IPv4 traffic to iptables' chains.
1693 bridge-nf-call-ip6tables - BOOLEAN
1694 1 : pass bridged IPv6 traffic to ip6tables' chains.
1698 bridge-nf-filter-vlan-tagged - BOOLEAN
1699 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1703 bridge-nf-filter-pppoe-tagged - BOOLEAN
1704 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1708 bridge-nf-pass-vlan-input-dev - BOOLEAN
1709 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1710 interface on the bridge and set the netfilter input device to the vlan.
1711 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1712 target work with vlan-on-top-of-bridge interfaces. When no matching
1713 vlan interface is found, or this switch is off, the input device is
1714 set to the bridge interface.
1715 0: disable bridge netfilter vlan interface lookup.
1718 proc/sys/net/sctp/* Variables:
1720 addip_enable - BOOLEAN
1721 Enable or disable extension of Dynamic Address Reconfiguration
1722 (ADD-IP) functionality specified in RFC5061. This extension provides
1723 the ability to dynamically add and remove new addresses for the SCTP
1726 1: Enable extension.
1728 0: Disable extension.
1732 addip_noauth_enable - BOOLEAN
1733 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1734 authentication to protect the operations of adding or removing new
1735 addresses. This requirement is mandated so that unauthorized hosts
1736 would not be able to hijack associations. However, older
1737 implementations may not have implemented this requirement while
1738 allowing the ADD-IP extension. For reasons of interoperability,
1739 we provide this variable to control the enforcement of the
1740 authentication requirement.
1742 1: Allow ADD-IP extension to be used without authentication. This
1743 should only be set in a closed environment for interoperability
1744 with older implementations.
1746 0: Enforce the authentication requirement
1750 auth_enable - BOOLEAN
1751 Enable or disable Authenticated Chunks extension. This extension
1752 provides the ability to send and receive authenticated chunks and is
1753 required for secure operation of Dynamic Address Reconfiguration
1756 1: Enable this extension.
1757 0: Disable this extension.
1761 prsctp_enable - BOOLEAN
1762 Enable or disable the Partial Reliability extension (RFC3758) which
1763 is used to notify peers that a given DATA should no longer be expected.
1771 The limit of the number of new packets that can be initially sent. It
1772 controls how bursty the generated traffic can be.
1776 association_max_retrans - INTEGER
1777 Set the maximum number for retransmissions that an association can
1778 attempt deciding that the remote end is unreachable. If this value
1779 is exceeded, the association is terminated.
1783 max_init_retransmits - INTEGER
1784 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1785 that an association will attempt before declaring the destination
1786 unreachable and terminating.
1790 path_max_retrans - INTEGER
1791 The maximum number of retransmissions that will be attempted on a given
1792 path. Once this threshold is exceeded, the path is considered
1793 unreachable, and new traffic will use a different path when the
1794 association is multihomed.
1798 pf_retrans - INTEGER
1799 The number of retransmissions that will be attempted on a given path
1800 before traffic is redirected to an alternate transport (should one
1801 exist). Note this is distinct from path_max_retrans, as a path that
1802 passes the pf_retrans threshold can still be used. Its only
1803 deprioritized when a transmission path is selected by the stack. This
1804 setting is primarily used to enable fast failover mechanisms without
1805 having to reduce path_max_retrans to a very low value. See:
1806 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1807 for details. Note also that a value of pf_retrans > path_max_retrans
1808 disables this feature
1812 rto_initial - INTEGER
1813 The initial round trip timeout value in milliseconds that will be used
1814 in calculating round trip times. This is the initial time interval
1815 for retransmissions.
1820 The maximum value (in milliseconds) of the round trip timeout. This
1821 is the largest time interval that can elapse between retransmissions.
1826 The minimum value (in milliseconds) of the round trip timeout. This
1827 is the smallest time interval the can elapse between retransmissions.
1831 hb_interval - INTEGER
1832 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1833 are sent at the specified interval on idle paths to probe the state of
1834 a given path between 2 associations.
1838 sack_timeout - INTEGER
1839 The amount of time (in milliseconds) that the implementation will wait
1844 valid_cookie_life - INTEGER
1845 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1846 is used during association establishment.
1850 cookie_preserve_enable - BOOLEAN
1851 Enable or disable the ability to extend the lifetime of the SCTP cookie
1852 that is used during the establishment phase of SCTP association
1854 1: Enable cookie lifetime extension.
1859 cookie_hmac_alg - STRING
1860 Select the hmac algorithm used when generating the cookie value sent by
1861 a listening sctp socket to a connecting client in the INIT-ACK chunk.
1866 Ability to assign md5 or sha1 as the selected alg is predicated on the
1867 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
1868 CONFIG_CRYPTO_SHA1).
1870 Default: Dependent on configuration. MD5 if available, else SHA1 if
1871 available, else none.
1873 rcvbuf_policy - INTEGER
1874 Determines if the receive buffer is attributed to the socket or to
1875 association. SCTP supports the capability to create multiple
1876 associations on a single socket. When using this capability, it is
1877 possible that a single stalled association that's buffering a lot
1878 of data may block other associations from delivering their data by
1879 consuming all of the receive buffer space. To work around this,
1880 the rcvbuf_policy could be set to attribute the receiver buffer space
1881 to each association instead of the socket. This prevents the described
1884 1: rcvbuf space is per association
1885 0: rcvbuf space is per socket
1889 sndbuf_policy - INTEGER
1890 Similar to rcvbuf_policy above, this applies to send buffer space.
1892 1: Send buffer is tracked per association
1893 0: Send buffer is tracked per socket.
1897 sctp_mem - vector of 3 INTEGERs: min, pressure, max
1898 Number of pages allowed for queueing by all SCTP sockets.
1900 min: Below this number of pages SCTP is not bothered about its
1901 memory appetite. When amount of memory allocated by SCTP exceeds
1902 this number, SCTP starts to moderate memory usage.
1904 pressure: This value was introduced to follow format of tcp_mem.
1906 max: Number of pages allowed for queueing by all SCTP sockets.
1908 Default is calculated at boot time from amount of available memory.
1910 sctp_rmem - vector of 3 INTEGERs: min, default, max
1911 Only the first value ("min") is used, "default" and "max" are
1914 min: Minimal size of receive buffer used by SCTP socket.
1915 It is guaranteed to each SCTP socket (but not association) even
1916 under moderate memory pressure.
1920 sctp_wmem - vector of 3 INTEGERs: min, default, max
1921 Currently this tunable has no effect.
1923 addr_scope_policy - INTEGER
1924 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1926 0 - Disable IPv4 address scoping
1927 1 - Enable IPv4 address scoping
1928 2 - Follow draft but allow IPv4 private addresses
1929 3 - Follow draft but allow IPv4 link local addresses
1934 /proc/sys/net/core/*
1935 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
1938 /proc/sys/net/unix/*
1939 max_dgram_qlen - INTEGER
1940 The maximum length of dgram socket receive queue
1947 /proc/sys/net/irda/*
1948 fast_poll_increase FIXME
1949 warn_noreply_time FIXME
1950 discovery_slots FIXME
1953 discovery_timeout FIXME
1954 lap_keepalive_time FIXME
1955 max_noreply_time FIXME
1956 max_tx_data_size FIXME
1958 min_tx_turn_time FIXME