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 - INTEGER
116 Maximum memory used to reassemble IP fragments. When
117 ipfrag_high_thresh bytes of memory is allocated for this purpose,
118 the fragment handler will toss packets until ipfrag_low_thresh
119 is reached. This also serves as a maximum limit to namespaces
120 different from the initial one.
122 ipfrag_low_thresh - INTEGER
123 Maximum memory used to reassemble IP fragments before the kernel
124 begins to remove incomplete fragment queues to free up resources.
125 The kernel still accepts new fragments for defragmentation.
127 ipfrag_time - INTEGER
128 Time in seconds to keep an IP fragment in memory.
130 ipfrag_max_dist - INTEGER
131 ipfrag_max_dist is a non-negative integer value which defines the
132 maximum "disorder" which is allowed among fragments which share a
133 common IP source address. Note that reordering of packets is
134 not unusual, but if a large number of fragments arrive from a source
135 IP address while a particular fragment queue remains incomplete, it
136 probably indicates that one or more fragments belonging to that queue
137 have been lost. When ipfrag_max_dist is positive, an additional check
138 is done on fragments before they are added to a reassembly queue - if
139 ipfrag_max_dist (or more) fragments have arrived from a particular IP
140 address between additions to any IP fragment queue using that source
141 address, it's presumed that one or more fragments in the queue are
142 lost. The existing fragment queue will be dropped, and a new one
143 started. An ipfrag_max_dist value of zero disables this check.
145 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
146 result in unnecessarily dropping fragment queues when normal
147 reordering of packets occurs, which could lead to poor application
148 performance. Using a very large value, e.g. 50000, increases the
149 likelihood of incorrectly reassembling IP fragments that originate
150 from different IP datagrams, which could result in data corruption.
155 inet_peer_threshold - INTEGER
156 The approximate size of the storage. Starting from this threshold
157 entries will be thrown aggressively. This threshold also determines
158 entries' time-to-live and time intervals between garbage collection
159 passes. More entries, less time-to-live, less GC interval.
161 inet_peer_minttl - INTEGER
162 Minimum time-to-live of entries. Should be enough to cover fragment
163 time-to-live on the reassembling side. This minimum time-to-live is
164 guaranteed if the pool size is less than inet_peer_threshold.
167 inet_peer_maxttl - INTEGER
168 Maximum time-to-live of entries. Unused entries will expire after
169 this period of time if there is no memory pressure on the pool (i.e.
170 when the number of entries in the pool is very small).
176 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
177 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
180 tcp_abort_on_overflow - BOOLEAN
181 If listening service is too slow to accept new connections,
182 reset them. Default state is FALSE. It means that if overflow
183 occurred due to a burst, connection will recover. Enable this
184 option _only_ if you are really sure that listening daemon
185 cannot be tuned to accept connections faster. Enabling this
186 option can harm clients of your server.
188 tcp_adv_win_scale - INTEGER
189 Count buffering overhead as bytes/2^tcp_adv_win_scale
190 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
192 Possible values are [-31, 31], inclusive.
195 tcp_allowed_congestion_control - STRING
196 Show/set the congestion control choices available to non-privileged
197 processes. The list is a subset of those listed in
198 tcp_available_congestion_control.
199 Default is "reno" and the default setting (tcp_congestion_control).
201 tcp_app_win - INTEGER
202 Reserve max(window/2^tcp_app_win, mss) of window for application
203 buffer. Value 0 is special, it means that nothing is reserved.
206 tcp_autocorking - BOOLEAN
207 Enable TCP auto corking :
208 When applications do consecutive small write()/sendmsg() system calls,
209 we try to coalesce these small writes as much as possible, to lower
210 total amount of sent packets. This is done if at least one prior
211 packet for the flow is waiting in Qdisc queues or device transmit
212 queue. Applications can still use TCP_CORK for optimal behavior
213 when they know how/when to uncork their sockets.
216 tcp_available_congestion_control - STRING
217 Shows the available congestion control choices that are registered.
218 More congestion control algorithms may be available as modules,
221 tcp_base_mss - INTEGER
222 The initial value of search_low to be used by the packetization layer
223 Path MTU discovery (MTU probing). If MTU probing is enabled,
224 this is the initial MSS used by the connection.
226 tcp_congestion_control - STRING
227 Set the congestion control algorithm to be used for new
228 connections. The algorithm "reno" is always available, but
229 additional choices may be available based on kernel configuration.
230 Default is set as part of kernel configuration.
231 For passive connections, the listener congestion control choice
233 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
236 Allows TCP to send "duplicate" SACKs.
238 tcp_early_retrans - INTEGER
239 Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
240 for triggering fast retransmit when the amount of outstanding data is
241 small and when no previously unsent data can be transmitted (such
242 that limited transmit could be used). Also controls the use of
243 Tail loss probe (TLP) that converts RTOs occurring due to tail
244 losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
248 2 enables ER but delays fast recovery and fast retransmit
249 by a fourth of RTT. This mitigates connection falsely
250 recovers when network has a small degree of reordering
251 (less than 3 packets).
252 3 enables delayed ER and TLP.
257 Control use of Explicit Congestion Notification (ECN) by TCP.
258 ECN is used only when both ends of the TCP connection indicate
259 support for it. This feature is useful in avoiding losses due
260 to congestion by allowing supporting routers to signal
261 congestion before having to drop packets.
263 0 Disable ECN. Neither initiate nor accept ECN.
264 1 Enable ECN when requested by incoming connections and
265 also request ECN on outgoing connection attempts.
266 2 Enable ECN when requested by incoming connections
267 but do not request ECN on outgoing connections.
270 tcp_ecn_fallback - BOOLEAN
271 If the kernel detects that ECN connection misbehaves, enable fall
272 back to non-ECN. Currently, this knob implements the fallback
273 from RFC3168, section 6.1.1.1., but we reserve that in future,
274 additional detection mechanisms could be implemented under this
275 knob. The value is not used, if tcp_ecn or per route (or congestion
276 control) ECN settings are disabled.
277 Default: 1 (fallback enabled)
280 Enable FACK congestion avoidance and fast retransmission.
281 The value is not used, if tcp_sack is not enabled.
283 tcp_fin_timeout - INTEGER
284 The length of time an orphaned (no longer referenced by any
285 application) connection will remain in the FIN_WAIT_2 state
286 before it is aborted at the local end. While a perfectly
287 valid "receive only" state for an un-orphaned connection, an
288 orphaned connection in FIN_WAIT_2 state could otherwise wait
289 forever for the remote to close its end of the connection.
294 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
295 F-RTO is an enhanced recovery algorithm for TCP retransmission
296 timeouts. It is particularly beneficial in networks where the
297 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
298 modification. It does not require any support from the peer.
300 By default it's enabled with a non-zero value. 0 disables F-RTO.
302 tcp_invalid_ratelimit - INTEGER
303 Limit the maximal rate for sending duplicate acknowledgments
304 in response to incoming TCP packets that are for an existing
305 connection but that are invalid due to any of these reasons:
307 (a) out-of-window sequence number,
308 (b) out-of-window acknowledgment number, or
309 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
311 This can help mitigate simple "ack loop" DoS attacks, wherein
312 a buggy or malicious middlebox or man-in-the-middle can
313 rewrite TCP header fields in manner that causes each endpoint
314 to think that the other is sending invalid TCP segments, thus
315 causing each side to send an unterminating stream of duplicate
316 acknowledgments for invalid segments.
318 Using 0 disables rate-limiting of dupacks in response to
319 invalid segments; otherwise this value specifies the minimal
320 space between sending such dupacks, in milliseconds.
322 Default: 500 (milliseconds).
324 tcp_keepalive_time - INTEGER
325 How often TCP sends out keepalive messages when keepalive is enabled.
328 tcp_keepalive_probes - INTEGER
329 How many keepalive probes TCP sends out, until it decides that the
330 connection is broken. Default value: 9.
332 tcp_keepalive_intvl - INTEGER
333 How frequently the probes are send out. Multiplied by
334 tcp_keepalive_probes it is time to kill not responding connection,
335 after probes started. Default value: 75sec i.e. connection
336 will be aborted after ~11 minutes of retries.
338 tcp_low_latency - BOOLEAN
339 If set, the TCP stack makes decisions that prefer lower
340 latency as opposed to higher throughput. By default, this
341 option is not set meaning that higher throughput is preferred.
342 An example of an application where this default should be
343 changed would be a Beowulf compute cluster.
346 tcp_max_orphans - INTEGER
347 Maximal number of TCP sockets not attached to any user file handle,
348 held by system. If this number is exceeded orphaned connections are
349 reset immediately and warning is printed. This limit exists
350 only to prevent simple DoS attacks, you _must_ not rely on this
351 or lower the limit artificially, but rather increase it
352 (probably, after increasing installed memory),
353 if network conditions require more than default value,
354 and tune network services to linger and kill such states
355 more aggressively. Let me to remind again: each orphan eats
356 up to ~64K of unswappable memory.
358 tcp_max_syn_backlog - INTEGER
359 Maximal number of remembered connection requests, which have not
360 received an acknowledgment from connecting client.
361 The minimal value is 128 for low memory machines, and it will
362 increase in proportion to the memory of machine.
363 If server suffers from overload, try increasing this number.
365 tcp_max_tw_buckets - INTEGER
366 Maximal number of timewait sockets held by system simultaneously.
367 If this number is exceeded time-wait socket is immediately destroyed
368 and warning is printed. This limit exists only to prevent
369 simple DoS attacks, you _must_ not lower the limit artificially,
370 but rather increase it (probably, after increasing installed memory),
371 if network conditions require more than default value.
373 tcp_mem - vector of 3 INTEGERs: min, pressure, max
374 min: below this number of pages TCP is not bothered about its
377 pressure: when amount of memory allocated by TCP exceeds this number
378 of pages, TCP moderates its memory consumption and enters memory
379 pressure mode, which is exited when memory consumption falls
382 max: number of pages allowed for queueing by all TCP sockets.
384 Defaults are calculated at boot time from amount of available
387 tcp_moderate_rcvbuf - BOOLEAN
388 If set, TCP performs receive buffer auto-tuning, attempting to
389 automatically size the buffer (no greater than tcp_rmem[2]) to
390 match the size required by the path for full throughput. Enabled by
393 tcp_mtu_probing - INTEGER
394 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
397 1 - Disabled by default, enabled when an ICMP black hole detected
398 2 - Always enabled, use initial MSS of tcp_base_mss.
400 tcp_probe_interval - INTEGER
401 Controls how often to start TCP Packetization-Layer Path MTU
402 Discovery reprobe. The default is reprobing every 10 minutes as
405 tcp_probe_threshold - INTEGER
406 Controls when TCP Packetization-Layer Path MTU Discovery probing
407 will stop in respect to the width of search range in bytes. Default
410 tcp_no_metrics_save - BOOLEAN
411 By default, TCP saves various connection metrics in the route cache
412 when the connection closes, so that connections established in the
413 near future can use these to set initial conditions. Usually, this
414 increases overall performance, but may sometimes cause performance
415 degradation. If set, TCP will not cache metrics on closing
418 tcp_orphan_retries - INTEGER
419 This value influences the timeout of a locally closed TCP connection,
420 when RTO retransmissions remain unacknowledged.
421 See tcp_retries2 for more details.
423 The default value is 8.
424 If your machine is a loaded WEB server,
425 you should think about lowering this value, such sockets
426 may consume significant resources. Cf. tcp_max_orphans.
428 tcp_reordering - INTEGER
429 Initial reordering level of packets in a TCP stream.
430 TCP stack can then dynamically adjust flow reordering level
431 between this initial value and tcp_max_reordering
434 tcp_max_reordering - INTEGER
435 Maximal reordering level of packets in a TCP stream.
436 300 is a fairly conservative value, but you might increase it
437 if paths are using per packet load balancing (like bonding rr mode)
440 tcp_retrans_collapse - BOOLEAN
441 Bug-to-bug compatibility with some broken printers.
442 On retransmit try to send bigger packets to work around bugs in
445 tcp_retries1 - INTEGER
446 This value influences the time, after which TCP decides, that
447 something is wrong due to unacknowledged RTO retransmissions,
448 and reports this suspicion to the network layer.
449 See tcp_retries2 for more details.
451 RFC 1122 recommends at least 3 retransmissions, which is the
454 tcp_retries2 - INTEGER
455 This value influences the timeout of an alive TCP connection,
456 when RTO retransmissions remain unacknowledged.
457 Given a value of N, a hypothetical TCP connection following
458 exponential backoff with an initial RTO of TCP_RTO_MIN would
459 retransmit N times before killing the connection at the (N+1)th RTO.
461 The default value of 15 yields a hypothetical timeout of 924.6
462 seconds and is a lower bound for the effective timeout.
463 TCP will effectively time out at the first RTO which exceeds the
464 hypothetical timeout.
466 RFC 1122 recommends at least 100 seconds for the timeout,
467 which corresponds to a value of at least 8.
469 tcp_rfc1337 - BOOLEAN
470 If set, the TCP stack behaves conforming to RFC1337. If unset,
471 we are not conforming to RFC, but prevent TCP TIME_WAIT
475 tcp_rmem - vector of 3 INTEGERs: min, default, max
476 min: Minimal size of receive buffer used by TCP sockets.
477 It is guaranteed to each TCP socket, even under moderate memory
481 default: initial size of receive buffer used by TCP sockets.
482 This value overrides net.core.rmem_default used by other protocols.
483 Default: 87380 bytes. This value results in window of 65535 with
484 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
485 less for default tcp_app_win. See below about these variables.
487 max: maximal size of receive buffer allowed for automatically
488 selected receiver buffers for TCP socket. This value does not override
489 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
490 automatic tuning of that socket's receive buffer size, in which
491 case this value is ignored.
492 Default: between 87380B and 6MB, depending on RAM size.
495 Enable select acknowledgments (SACKS).
497 tcp_slow_start_after_idle - BOOLEAN
498 If set, provide RFC2861 behavior and time out the congestion
499 window after an idle period. An idle period is defined at
500 the current RTO. If unset, the congestion window will not
501 be timed out after an idle period.
505 Use the Host requirements interpretation of the TCP urgent pointer field.
506 Most hosts use the older BSD interpretation, so if you turn this on
507 Linux might not communicate correctly with them.
510 tcp_synack_retries - INTEGER
511 Number of times SYNACKs for a passive TCP connection attempt will
512 be retransmitted. Should not be higher than 255. Default value
513 is 5, which corresponds to 31seconds till the last retransmission
514 with the current initial RTO of 1second. With this the final timeout
515 for a passive TCP connection will happen after 63seconds.
517 tcp_syncookies - BOOLEAN
518 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
519 Send out syncookies when the syn backlog queue of a socket
520 overflows. This is to prevent against the common 'SYN flood attack'
523 Note, that syncookies is fallback facility.
524 It MUST NOT be used to help highly loaded servers to stand
525 against legal connection rate. If you see SYN flood warnings
526 in your logs, but investigation shows that they occur
527 because of overload with legal connections, you should tune
528 another parameters until this warning disappear.
529 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
531 syncookies seriously violate TCP protocol, do not allow
532 to use TCP extensions, can result in serious degradation
533 of some services (f.e. SMTP relaying), visible not by you,
534 but your clients and relays, contacting you. While you see
535 SYN flood warnings in logs not being really flooded, your server
536 is seriously misconfigured.
538 If you want to test which effects syncookies have to your
539 network connections you can set this knob to 2 to enable
540 unconditionally generation of syncookies.
542 tcp_fastopen - INTEGER
543 Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
544 in the opening SYN packet. To use this feature, the client application
545 must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
546 connect() to perform a TCP handshake automatically.
548 The values (bitmap) are
549 1: Enables sending data in the opening SYN on the client w/ MSG_FASTOPEN.
550 2: Enables TCP Fast Open on the server side, i.e., allowing data in
551 a SYN packet to be accepted and passed to the application before
552 3-way hand shake finishes.
553 4: Send data in the opening SYN regardless of cookie availability and
554 without a cookie option.
555 0x100: Accept SYN data w/o validating the cookie.
556 0x200: Accept data-in-SYN w/o any cookie option present.
557 0x400/0x800: Enable Fast Open on all listeners regardless of the
558 TCP_FASTOPEN socket option. The two different flags designate two
559 different ways of setting max_qlen without the TCP_FASTOPEN socket
564 Note that the client & server side Fast Open flags (1 and 2
565 respectively) must be also enabled before the rest of flags can take
568 See include/net/tcp.h and the code for more details.
570 tcp_syn_retries - INTEGER
571 Number of times initial SYNs for an active TCP connection attempt
572 will be retransmitted. Should not be higher than 255. Default value
573 is 6, which corresponds to 63seconds till the last retransmission
574 with the current initial RTO of 1second. With this the final timeout
575 for an active TCP connection attempt will happen after 127seconds.
577 tcp_timestamps - BOOLEAN
578 Enable timestamps as defined in RFC1323.
580 tcp_min_tso_segs - INTEGER
581 Minimal number of segments per TSO frame.
582 Since linux-3.12, TCP does an automatic sizing of TSO frames,
583 depending on flow rate, instead of filling 64Kbytes packets.
584 For specific usages, it's possible to force TCP to build big
585 TSO frames. Note that TCP stack might split too big TSO packets
586 if available window is too small.
589 tcp_pacing_ss_ratio - INTEGER
590 sk->sk_pacing_rate is set by TCP stack using a ratio applied
591 to current rate. (current_rate = cwnd * mss / srtt)
592 If TCP is in slow start, tcp_pacing_ss_ratio is applied
593 to let TCP probe for bigger speeds, assuming cwnd can be
594 doubled every other RTT.
597 tcp_pacing_ca_ratio - INTEGER
598 sk->sk_pacing_rate is set by TCP stack using a ratio applied
599 to current rate. (current_rate = cwnd * mss / srtt)
600 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
601 is applied to conservatively probe for bigger throughput.
604 tcp_tso_win_divisor - INTEGER
605 This allows control over what percentage of the congestion window
606 can be consumed by a single TSO frame.
607 The setting of this parameter is a choice between burstiness and
608 building larger TSO frames.
611 tcp_tw_recycle - BOOLEAN
612 Enable fast recycling TIME-WAIT sockets. Default value is 0.
613 It should not be changed without advice/request of technical
616 tcp_tw_reuse - BOOLEAN
617 Allow to reuse TIME-WAIT sockets for new connections when it is
618 safe from protocol viewpoint. Default value is 0.
619 It should not be changed without advice/request of technical
622 tcp_window_scaling - BOOLEAN
623 Enable window scaling as defined in RFC1323.
625 tcp_wmem - vector of 3 INTEGERs: min, default, max
626 min: Amount of memory reserved for send buffers for TCP sockets.
627 Each TCP socket has rights to use it due to fact of its birth.
630 default: initial size of send buffer used by TCP sockets. This
631 value overrides net.core.wmem_default used by other protocols.
632 It is usually lower than net.core.wmem_default.
635 max: Maximal amount of memory allowed for automatically tuned
636 send buffers for TCP sockets. This value does not override
637 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
638 automatic tuning of that socket's send buffer size, in which case
639 this value is ignored.
640 Default: between 64K and 4MB, depending on RAM size.
642 tcp_notsent_lowat - UNSIGNED INTEGER
643 A TCP socket can control the amount of unsent bytes in its write queue,
644 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
645 reports POLLOUT events if the amount of unsent bytes is below a per
646 socket value, and if the write queue is not full. sendmsg() will
647 also not add new buffers if the limit is hit.
649 This global variable controls the amount of unsent data for
650 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
651 to the global variable has immediate effect.
653 Default: UINT_MAX (0xFFFFFFFF)
655 tcp_workaround_signed_windows - BOOLEAN
656 If set, assume no receipt of a window scaling option means the
657 remote TCP is broken and treats the window as a signed quantity.
658 If unset, assume the remote TCP is not broken even if we do
659 not receive a window scaling option from them.
662 tcp_thin_linear_timeouts - BOOLEAN
663 Enable dynamic triggering of linear timeouts for thin streams.
664 If set, a check is performed upon retransmission by timeout to
665 determine if the stream is thin (less than 4 packets in flight).
666 As long as the stream is found to be thin, up to 6 linear
667 timeouts may be performed before exponential backoff mode is
668 initiated. This improves retransmission latency for
669 non-aggressive thin streams, often found to be time-dependent.
670 For more information on thin streams, see
671 Documentation/networking/tcp-thin.txt
674 tcp_thin_dupack - BOOLEAN
675 Enable dynamic triggering of retransmissions after one dupACK
676 for thin streams. If set, a check is performed upon reception
677 of a dupACK to determine if the stream is thin (less than 4
678 packets in flight). As long as the stream is found to be thin,
679 data is retransmitted on the first received dupACK. This
680 improves retransmission latency for non-aggressive thin
681 streams, often found to be time-dependent.
682 For more information on thin streams, see
683 Documentation/networking/tcp-thin.txt
686 tcp_limit_output_bytes - INTEGER
687 Controls TCP Small Queue limit per tcp socket.
688 TCP bulk sender tends to increase packets in flight until it
689 gets losses notifications. With SNDBUF autotuning, this can
690 result in a large amount of packets queued in qdisc/device
691 on the local machine, hurting latency of other flows, for
692 typical pfifo_fast qdiscs.
693 tcp_limit_output_bytes limits the number of bytes on qdisc
694 or device to reduce artificial RTT/cwnd and reduce bufferbloat.
697 tcp_challenge_ack_limit - INTEGER
698 Limits number of Challenge ACK sent per second, as recommended
699 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
704 udp_mem - vector of 3 INTEGERs: min, pressure, max
705 Number of pages allowed for queueing by all UDP sockets.
707 min: Below this number of pages UDP is not bothered about its
708 memory appetite. When amount of memory allocated by UDP exceeds
709 this number, UDP starts to moderate memory usage.
711 pressure: This value was introduced to follow format of tcp_mem.
713 max: Number of pages allowed for queueing by all UDP sockets.
715 Default is calculated at boot time from amount of available memory.
717 udp_rmem_min - INTEGER
718 Minimal size of receive buffer used by UDP sockets in moderation.
719 Each UDP socket is able to use the size for receiving data, even if
720 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
723 udp_wmem_min - INTEGER
724 Minimal size of send buffer used by UDP sockets in moderation.
725 Each UDP socket is able to use the size for sending data, even if
726 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
731 cipso_cache_enable - BOOLEAN
732 If set, enable additions to and lookups from the CIPSO label mapping
733 cache. If unset, additions are ignored and lookups always result in a
734 miss. However, regardless of the setting the cache is still
735 invalidated when required when means you can safely toggle this on and
736 off and the cache will always be "safe".
739 cipso_cache_bucket_size - INTEGER
740 The CIPSO label cache consists of a fixed size hash table with each
741 hash bucket containing a number of cache entries. This variable limits
742 the number of entries in each hash bucket; the larger the value the
743 more CIPSO label mappings that can be cached. When the number of
744 entries in a given hash bucket reaches this limit adding new entries
745 causes the oldest entry in the bucket to be removed to make room.
748 cipso_rbm_optfmt - BOOLEAN
749 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
750 the CIPSO draft specification (see Documentation/netlabel for details).
751 This means that when set the CIPSO tag will be padded with empty
752 categories in order to make the packet data 32-bit aligned.
755 cipso_rbm_structvalid - BOOLEAN
756 If set, do a very strict check of the CIPSO option when
757 ip_options_compile() is called. If unset, relax the checks done during
758 ip_options_compile(). Either way is "safe" as errors are caught else
759 where in the CIPSO processing code but setting this to 0 (False) should
760 result in less work (i.e. it should be faster) but could cause problems
761 with other implementations that require strict checking.
766 ip_local_port_range - 2 INTEGERS
767 Defines the local port range that is used by TCP and UDP to
768 choose the local port. The first number is the first, the
769 second the last local port number.
770 If possible, it is better these numbers have different parity.
771 (one even and one odd values)
772 The default values are 32768 and 60999 respectively.
774 ip_local_reserved_ports - list of comma separated ranges
775 Specify the ports which are reserved for known third-party
776 applications. These ports will not be used by automatic port
777 assignments (e.g. when calling connect() or bind() with port
778 number 0). Explicit port allocation behavior is unchanged.
780 The format used for both input and output is a comma separated
781 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
782 10). Writing to the file will clear all previously reserved
783 ports and update the current list with the one given in the
786 Note that ip_local_port_range and ip_local_reserved_ports
787 settings are independent and both are considered by the kernel
788 when determining which ports are available for automatic port
791 You can reserve ports which are not in the current
792 ip_local_port_range, e.g.:
794 $ cat /proc/sys/net/ipv4/ip_local_port_range
796 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
799 although this is redundant. However such a setting is useful
800 if later the port range is changed to a value that will
801 include the reserved ports.
805 ip_nonlocal_bind - BOOLEAN
806 If set, allows processes to bind() to non-local IP addresses,
807 which can be quite useful - but may break some applications.
811 If set non-zero, enables support for dynamic addresses.
812 If set to a non-zero value larger than 1, a kernel log
813 message will be printed when dynamic address rewriting
817 ip_early_demux - BOOLEAN
818 Optimize input packet processing down to one demux for
819 certain kinds of local sockets. Currently we only do this
820 for established TCP sockets.
822 It may add an additional cost for pure routing workloads that
823 reduces overall throughput, in such case you should disable it.
826 icmp_echo_ignore_all - BOOLEAN
827 If set non-zero, then the kernel will ignore all ICMP ECHO
831 icmp_echo_ignore_broadcasts - BOOLEAN
832 If set non-zero, then the kernel will ignore all ICMP ECHO and
833 TIMESTAMP requests sent to it via broadcast/multicast.
836 icmp_ratelimit - INTEGER
837 Limit the maximal rates for sending ICMP packets whose type matches
838 icmp_ratemask (see below) to specific targets.
839 0 to disable any limiting,
840 otherwise the minimal space between responses in milliseconds.
841 Note that another sysctl, icmp_msgs_per_sec limits the number
842 of ICMP packets sent on all targets.
845 icmp_msgs_per_sec - INTEGER
846 Limit maximal number of ICMP packets sent per second from this host.
847 Only messages whose type matches icmp_ratemask (see below) are
848 controlled by this limit.
851 icmp_msgs_burst - INTEGER
852 icmp_msgs_per_sec controls number of ICMP packets sent per second,
853 while icmp_msgs_burst controls the burst size of these packets.
856 icmp_ratemask - INTEGER
857 Mask made of ICMP types for which rates are being limited.
858 Significant bits: IHGFEDCBA9876543210
859 Default mask: 0000001100000011000 (6168)
861 Bit definitions (see include/linux/icmp.h):
863 3 Destination Unreachable *
868 C Parameter Problem *
873 H Address Mask Request
876 * These are rate limited by default (see default mask above)
878 icmp_ignore_bogus_error_responses - BOOLEAN
879 Some routers violate RFC1122 by sending bogus responses to broadcast
880 frames. Such violations are normally logged via a kernel warning.
881 If this is set to TRUE, the kernel will not give such warnings, which
882 will avoid log file clutter.
885 icmp_errors_use_inbound_ifaddr - BOOLEAN
887 If zero, icmp error messages except redirects are sent with the primary
888 address of the exiting interface.
890 If non-zero, the message will be sent with the primary address of
891 the interface that received the packet that caused the icmp error.
892 This is the behaviour network many administrators will expect from
893 a router. And it can make debugging complicated network layouts
896 Note that if no primary address exists for the interface selected,
897 then the primary address of the first non-loopback interface that
898 has one will be used regardless of this setting.
900 The source address selection of icmp redirect messages is controlled by
901 icmp_errors_use_inbound_ifaddr.
904 icmp_redirects_use_orig_daddr - BOOLEAN
906 If zero, icmp redirect messages are sent using the address specified for
907 other icmp errors by icmp_errors_use_inbound_ifaddr.
909 If non-zero, the message will be sent with the destination address of
910 the packet that caused the icmp redirect.
911 This behaviour is the preferred one on VRRP routers (see RFC 5798
917 igmp_max_memberships - INTEGER
918 Change the maximum number of multicast groups we can subscribe to.
921 Theoretical maximum value is bounded by having to send a membership
922 report in a single datagram (i.e. the report can't span multiple
923 datagrams, or risk confusing the switch and leaving groups you don't
926 The number of supported groups 'M' is bounded by the number of group
927 report entries you can fit into a single datagram of 65535 bytes.
929 M = 65536-sizeof (ip header)/(sizeof(Group record))
931 Group records are variable length, with a minimum of 12 bytes.
932 So net.ipv4.igmp_max_memberships should not be set higher than:
934 (65536-24) / 12 = 5459
936 The value 5459 assumes no IP header options, so in practice
937 this number may be lower.
939 conf/interface/* changes special settings per interface (where
940 "interface" is the name of your network interface)
942 conf/all/* is special, changes the settings for all interfaces
945 Controls the IGMP query robustness variable (see RFC2236 8.1).
946 Default: 2 (as specified by RFC2236 8.1)
947 Minimum: 1 (as specified by RFC6636 4.5)
949 log_martians - BOOLEAN
950 Log packets with impossible addresses to kernel log.
951 log_martians for the interface will be enabled if at least one of
952 conf/{all,interface}/log_martians is set to TRUE,
953 it will be disabled otherwise
955 accept_redirects - BOOLEAN
956 Accept ICMP redirect messages.
957 accept_redirects for the interface will be enabled if:
958 - both conf/{all,interface}/accept_redirects are TRUE in the case
959 forwarding for the interface is enabled
961 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
962 case forwarding for the interface is disabled
963 accept_redirects for the interface will be disabled otherwise
968 Enable IP forwarding on this interface.
970 mc_forwarding - BOOLEAN
971 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
972 and a multicast routing daemon is required.
973 conf/all/mc_forwarding must also be set to TRUE to enable multicast
974 routing for the interface
977 Integer value used to differentiate the devices by the medium they
978 are attached to. Two devices can have different id values when
979 the broadcast packets are received only on one of them.
980 The default value 0 means that the device is the only interface
981 to its medium, value of -1 means that medium is not known.
983 Currently, it is used to change the proxy_arp behavior:
984 the proxy_arp feature is enabled for packets forwarded between
985 two devices attached to different media.
989 proxy_arp for the interface will be enabled if at least one of
990 conf/{all,interface}/proxy_arp is set to TRUE,
991 it will be disabled otherwise
993 proxy_arp_pvlan - BOOLEAN
994 Private VLAN proxy arp.
995 Basically allow proxy arp replies back to the same interface
996 (from which the ARP request/solicitation was received).
998 This is done to support (ethernet) switch features, like RFC
999 3069, where the individual ports are NOT allowed to
1000 communicate with each other, but they are allowed to talk to
1001 the upstream router. As described in RFC 3069, it is possible
1002 to allow these hosts to communicate through the upstream
1003 router by proxy_arp'ing. Don't need to be used together with
1006 This technology is known by different names:
1007 In RFC 3069 it is called VLAN Aggregation.
1008 Cisco and Allied Telesyn call it Private VLAN.
1009 Hewlett-Packard call it Source-Port filtering or port-isolation.
1010 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1012 shared_media - BOOLEAN
1013 Send(router) or accept(host) RFC1620 shared media redirects.
1014 Overrides ip_secure_redirects.
1015 shared_media for the interface will be enabled if at least one of
1016 conf/{all,interface}/shared_media is set to TRUE,
1017 it will be disabled otherwise
1020 secure_redirects - BOOLEAN
1021 Accept ICMP redirect messages only for gateways,
1022 listed in default gateway list.
1023 secure_redirects for the interface will be enabled if at least one of
1024 conf/{all,interface}/secure_redirects is set to TRUE,
1025 it will be disabled otherwise
1028 send_redirects - BOOLEAN
1029 Send redirects, if router.
1030 send_redirects for the interface will be enabled if at least one of
1031 conf/{all,interface}/send_redirects is set to TRUE,
1032 it will be disabled otherwise
1035 bootp_relay - BOOLEAN
1036 Accept packets with source address 0.b.c.d destined
1037 not to this host as local ones. It is supposed, that
1038 BOOTP relay daemon will catch and forward such packets.
1039 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1042 Not Implemented Yet.
1044 accept_source_route - BOOLEAN
1045 Accept packets with SRR option.
1046 conf/all/accept_source_route must also be set to TRUE to accept packets
1047 with SRR option on the interface
1048 default TRUE (router)
1051 accept_local - BOOLEAN
1052 Accept packets with local source addresses. In combination with
1053 suitable routing, this can be used to direct packets between two
1054 local interfaces over the wire and have them accepted properly.
1057 route_localnet - BOOLEAN
1058 Do not consider loopback addresses as martian source or destination
1059 while routing. This enables the use of 127/8 for local routing purposes.
1063 0 - No source validation.
1064 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1065 Each incoming packet is tested against the FIB and if the interface
1066 is not the best reverse path the packet check will fail.
1067 By default failed packets are discarded.
1068 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1069 Each incoming packet's source address is also tested against the FIB
1070 and if the source address is not reachable via any interface
1071 the packet check will fail.
1073 Current recommended practice in RFC3704 is to enable strict mode
1074 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1075 or other complicated routing, then loose mode is recommended.
1077 The max value from conf/{all,interface}/rp_filter is used
1078 when doing source validation on the {interface}.
1080 Default value is 0. Note that some distributions enable it
1083 arp_filter - BOOLEAN
1084 1 - Allows you to have multiple network interfaces on the same
1085 subnet, and have the ARPs for each interface be answered
1086 based on whether or not the kernel would route a packet from
1087 the ARP'd IP out that interface (therefore you must use source
1088 based routing for this to work). In other words it allows control
1089 of which cards (usually 1) will respond to an arp request.
1091 0 - (default) The kernel can respond to arp requests with addresses
1092 from other interfaces. This may seem wrong but it usually makes
1093 sense, because it increases the chance of successful communication.
1094 IP addresses are owned by the complete host on Linux, not by
1095 particular interfaces. Only for more complex setups like load-
1096 balancing, does this behaviour cause problems.
1098 arp_filter for the interface will be enabled if at least one of
1099 conf/{all,interface}/arp_filter is set to TRUE,
1100 it will be disabled otherwise
1102 arp_announce - INTEGER
1103 Define different restriction levels for announcing the local
1104 source IP address from IP packets in ARP requests sent on
1106 0 - (default) Use any local address, configured on any interface
1107 1 - Try to avoid local addresses that are not in the target's
1108 subnet for this interface. This mode is useful when target
1109 hosts reachable via this interface require the source IP
1110 address in ARP requests to be part of their logical network
1111 configured on the receiving interface. When we generate the
1112 request we will check all our subnets that include the
1113 target IP and will preserve the source address if it is from
1114 such subnet. If there is no such subnet we select source
1115 address according to the rules for level 2.
1116 2 - Always use the best local address for this target.
1117 In this mode we ignore the source address in the IP packet
1118 and try to select local address that we prefer for talks with
1119 the target host. Such local address is selected by looking
1120 for primary IP addresses on all our subnets on the outgoing
1121 interface that include the target IP address. If no suitable
1122 local address is found we select the first local address
1123 we have on the outgoing interface or on all other interfaces,
1124 with the hope we will receive reply for our request and
1125 even sometimes no matter the source IP address we announce.
1127 The max value from conf/{all,interface}/arp_announce is used.
1129 Increasing the restriction level gives more chance for
1130 receiving answer from the resolved target while decreasing
1131 the level announces more valid sender's information.
1133 arp_ignore - INTEGER
1134 Define different modes for sending replies in response to
1135 received ARP requests that resolve local target IP addresses:
1136 0 - (default): reply for any local target IP address, configured
1138 1 - reply only if the target IP address is local address
1139 configured on the incoming interface
1140 2 - reply only if the target IP address is local address
1141 configured on the incoming interface and both with the
1142 sender's IP address are part from same subnet on this interface
1143 3 - do not reply for local addresses configured with scope host,
1144 only resolutions for global and link addresses are replied
1146 8 - do not reply for all local addresses
1148 The max value from conf/{all,interface}/arp_ignore is used
1149 when ARP request is received on the {interface}
1151 arp_notify - BOOLEAN
1152 Define mode for notification of address and device changes.
1153 0 - (default): do nothing
1154 1 - Generate gratuitous arp requests when device is brought up
1155 or hardware address changes.
1157 arp_accept - BOOLEAN
1158 Define behavior for gratuitous ARP frames who's IP is not
1159 already present in the ARP table:
1160 0 - don't create new entries in the ARP table
1161 1 - create new entries in the ARP table
1163 Both replies and requests type gratuitous arp will trigger the
1164 ARP table to be updated, if this setting is on.
1166 If the ARP table already contains the IP address of the
1167 gratuitous arp frame, the arp table will be updated regardless
1168 if this setting is on or off.
1170 mcast_solicit - INTEGER
1171 The maximum number of multicast probes in INCOMPLETE state,
1172 when the associated hardware address is unknown. Defaults
1175 ucast_solicit - INTEGER
1176 The maximum number of unicast probes in PROBE state, when
1177 the hardware address is being reconfirmed. Defaults to 3.
1179 app_solicit - INTEGER
1180 The maximum number of probes to send to the user space ARP daemon
1181 via netlink before dropping back to multicast probes (see
1182 mcast_resolicit). Defaults to 0.
1184 mcast_resolicit - INTEGER
1185 The maximum number of multicast probes after unicast and
1186 app probes in PROBE state. Defaults to 0.
1188 disable_policy - BOOLEAN
1189 Disable IPSEC policy (SPD) for this interface
1191 disable_xfrm - BOOLEAN
1192 Disable IPSEC encryption on this interface, whatever the policy
1194 igmpv2_unsolicited_report_interval - INTEGER
1195 The interval in milliseconds in which the next unsolicited
1196 IGMPv1 or IGMPv2 report retransmit will take place.
1197 Default: 10000 (10 seconds)
1199 igmpv3_unsolicited_report_interval - INTEGER
1200 The interval in milliseconds in which the next unsolicited
1201 IGMPv3 report retransmit will take place.
1202 Default: 1000 (1 seconds)
1204 promote_secondaries - BOOLEAN
1205 When a primary IP address is removed from this interface
1206 promote a corresponding secondary IP address instead of
1207 removing all the corresponding secondary IP addresses.
1211 Allows you to write a number, which can be used as required.
1214 xfrm4_gc_thresh - INTEGER
1215 The threshold at which we will start garbage collecting for IPv4
1216 destination cache entries. At twice this value the system will
1217 refuse new allocations.
1219 igmp_link_local_mcast_reports - BOOLEAN
1220 Enable IGMP reports for link local multicast groups in the
1225 kuznet@ms2.inr.ac.ru
1231 delon.nicolas@wanadoo.fr
1236 /proc/sys/net/ipv6/* Variables:
1238 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1239 apply to IPv6 [XXX?].
1241 bindv6only - BOOLEAN
1242 Default value for IPV6_V6ONLY socket option,
1243 which restricts use of the IPv6 socket to IPv6 communication
1245 TRUE: disable IPv4-mapped address feature
1246 FALSE: enable IPv4-mapped address feature
1248 Default: FALSE (as specified in RFC3493)
1250 flowlabel_consistency - BOOLEAN
1251 Protect the consistency (and unicity) of flow label.
1252 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1258 auto_flowlabels - INTEGER
1259 Automatically generate flow labels based on a flow hash of the
1260 packet. This allows intermediate devices, such as routers, to
1261 identify packet flows for mechanisms like Equal Cost Multipath
1262 Routing (see RFC 6438).
1263 0: automatic flow labels are completely disabled
1264 1: automatic flow labels are enabled by default, they can be
1265 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1267 2: automatic flow labels are allowed, they may be enabled on a
1268 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1269 3: automatic flow labels are enabled and enforced, they cannot
1270 be disabled by the socket option
1273 flowlabel_state_ranges - BOOLEAN
1274 Split the flow label number space into two ranges. 0-0x7FFFF is
1275 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1276 is reserved for stateless flow labels as described in RFC6437.
1281 anycast_src_echo_reply - BOOLEAN
1282 Controls the use of anycast addresses as source addresses for ICMPv6
1288 idgen_delay - INTEGER
1289 Controls the delay in seconds after which time to retry
1290 privacy stable address generation if a DAD conflict is
1292 Default: 1 (as specified in RFC7217)
1294 idgen_retries - INTEGER
1295 Controls the number of retries to generate a stable privacy
1296 address if a DAD conflict is detected.
1297 Default: 3 (as specified in RFC7217)
1300 Controls the MLD query robustness variable (see RFC3810 9.1).
1301 Default: 2 (as specified by RFC3810 9.1)
1302 Minimum: 1 (as specified by RFC6636 4.5)
1306 ip6frag_high_thresh - INTEGER
1307 Maximum memory used to reassemble IPv6 fragments. When
1308 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1309 the fragment handler will toss packets until ip6frag_low_thresh
1312 ip6frag_low_thresh - INTEGER
1313 See ip6frag_high_thresh
1315 ip6frag_time - INTEGER
1316 Time in seconds to keep an IPv6 fragment in memory.
1319 Change the interface-specific default settings.
1323 Change all the interface-specific settings.
1325 [XXX: Other special features than forwarding?]
1327 conf/all/forwarding - BOOLEAN
1328 Enable global IPv6 forwarding between all interfaces.
1330 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1331 to control which interfaces may forward packets and which not.
1333 This also sets all interfaces' Host/Router setting
1334 'forwarding' to the specified value. See below for details.
1336 This referred to as global forwarding.
1341 fwmark_reflect - BOOLEAN
1342 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1343 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1344 If unset, these packets have a fwmark of zero. If set, they have the
1345 fwmark of the packet they are replying to.
1349 Change special settings per interface.
1351 The functional behaviour for certain settings is different
1352 depending on whether local forwarding is enabled or not.
1355 Accept Router Advertisements; autoconfigure using them.
1357 It also determines whether or not to transmit Router
1358 Solicitations. If and only if the functional setting is to
1359 accept Router Advertisements, Router Solicitations will be
1362 Possible values are:
1363 0 Do not accept Router Advertisements.
1364 1 Accept Router Advertisements if forwarding is disabled.
1365 2 Overrule forwarding behaviour. Accept Router Advertisements
1366 even if forwarding is enabled.
1368 Functional default: enabled if local forwarding is disabled.
1369 disabled if local forwarding is enabled.
1371 accept_ra_defrtr - BOOLEAN
1372 Learn default router in Router Advertisement.
1374 Functional default: enabled if accept_ra is enabled.
1375 disabled if accept_ra is disabled.
1377 accept_ra_from_local - BOOLEAN
1378 Accept RA with source-address that is found on local machine
1379 if the RA is otherwise proper and able to be accepted.
1380 Default is to NOT accept these as it may be an un-intended
1384 enabled if accept_ra_from_local is enabled
1385 on a specific interface.
1386 disabled if accept_ra_from_local is disabled
1387 on a specific interface.
1389 accept_ra_min_hop_limit - INTEGER
1390 Minimum hop limit Information in Router Advertisement.
1392 Hop limit Information in Router Advertisement less than this
1393 variable shall be ignored.
1397 accept_ra_pinfo - BOOLEAN
1398 Learn Prefix Information in Router Advertisement.
1400 Functional default: enabled if accept_ra is enabled.
1401 disabled if accept_ra is disabled.
1403 accept_ra_rt_info_max_plen - INTEGER
1404 Maximum prefix length of Route Information in RA.
1406 Route Information w/ prefix larger than or equal to this
1407 variable shall be ignored.
1409 Functional default: 0 if accept_ra_rtr_pref is enabled.
1410 -1 if accept_ra_rtr_pref is disabled.
1412 accept_ra_rtr_pref - BOOLEAN
1413 Accept Router Preference in RA.
1415 Functional default: enabled if accept_ra is enabled.
1416 disabled if accept_ra is disabled.
1418 accept_ra_mtu - BOOLEAN
1419 Apply the MTU value specified in RA option 5 (RFC4861). If
1420 disabled, the MTU specified in the RA will be ignored.
1422 Functional default: enabled if accept_ra is enabled.
1423 disabled if accept_ra is disabled.
1425 accept_redirects - BOOLEAN
1428 Functional default: enabled if local forwarding is disabled.
1429 disabled if local forwarding is enabled.
1431 accept_source_route - INTEGER
1432 Accept source routing (routing extension header).
1434 >= 0: Accept only routing header type 2.
1435 < 0: Do not accept routing header.
1440 Autoconfigure addresses using Prefix Information in Router
1443 Functional default: enabled if accept_ra_pinfo is enabled.
1444 disabled if accept_ra_pinfo is disabled.
1446 dad_transmits - INTEGER
1447 The amount of Duplicate Address Detection probes to send.
1450 forwarding - INTEGER
1451 Configure interface-specific Host/Router behaviour.
1453 Note: It is recommended to have the same setting on all
1454 interfaces; mixed router/host scenarios are rather uncommon.
1456 Possible values are:
1457 0 Forwarding disabled
1458 1 Forwarding enabled
1462 By default, Host behaviour is assumed. This means:
1464 1. IsRouter flag is not set in Neighbour Advertisements.
1465 2. If accept_ra is TRUE (default), transmit Router
1467 3. If accept_ra is TRUE (default), accept Router
1468 Advertisements (and do autoconfiguration).
1469 4. If accept_redirects is TRUE (default), accept Redirects.
1473 If local forwarding is enabled, Router behaviour is assumed.
1474 This means exactly the reverse from the above:
1476 1. IsRouter flag is set in Neighbour Advertisements.
1477 2. Router Solicitations are not sent unless accept_ra is 2.
1478 3. Router Advertisements are ignored unless accept_ra is 2.
1479 4. Redirects are ignored.
1481 Default: 0 (disabled) if global forwarding is disabled (default),
1482 otherwise 1 (enabled).
1485 Default Hop Limit to set.
1489 Default Maximum Transfer Unit
1490 Default: 1280 (IPv6 required minimum)
1492 ip_nonlocal_bind - BOOLEAN
1493 If set, allows processes to bind() to non-local IPv6 addresses,
1494 which can be quite useful - but may break some applications.
1497 router_probe_interval - INTEGER
1498 Minimum interval (in seconds) between Router Probing described
1503 router_solicitation_delay - INTEGER
1504 Number of seconds to wait after interface is brought up
1505 before sending Router Solicitations.
1508 router_solicitation_interval - INTEGER
1509 Number of seconds to wait between Router Solicitations.
1512 router_solicitations - INTEGER
1513 Number of Router Solicitations to send until assuming no
1514 routers are present.
1517 use_oif_addrs_only - BOOLEAN
1518 When enabled, the candidate source addresses for destinations
1519 routed via this interface are restricted to the set of addresses
1520 configured on this interface (vis. RFC 6724, section 4).
1524 use_tempaddr - INTEGER
1525 Preference for Privacy Extensions (RFC3041).
1526 <= 0 : disable Privacy Extensions
1527 == 1 : enable Privacy Extensions, but prefer public
1528 addresses over temporary addresses.
1529 > 1 : enable Privacy Extensions and prefer temporary
1530 addresses over public addresses.
1531 Default: 0 (for most devices)
1532 -1 (for point-to-point devices and loopback devices)
1534 temp_valid_lft - INTEGER
1535 valid lifetime (in seconds) for temporary addresses.
1536 Default: 604800 (7 days)
1538 temp_prefered_lft - INTEGER
1539 Preferred lifetime (in seconds) for temporary addresses.
1540 Default: 86400 (1 day)
1542 max_desync_factor - INTEGER
1543 Maximum value for DESYNC_FACTOR, which is a random value
1544 that ensures that clients don't synchronize with each
1545 other and generate new addresses at exactly the same time.
1546 value is in seconds.
1549 regen_max_retry - INTEGER
1550 Number of attempts before give up attempting to generate
1551 valid temporary addresses.
1554 max_addresses - INTEGER
1555 Maximum number of autoconfigured addresses per interface. Setting
1556 to zero disables the limitation. It is not recommended to set this
1557 value too large (or to zero) because it would be an easy way to
1558 crash the kernel by allowing too many addresses to be created.
1561 disable_ipv6 - BOOLEAN
1562 Disable IPv6 operation. If accept_dad is set to 2, this value
1563 will be dynamically set to TRUE if DAD fails for the link-local
1565 Default: FALSE (enable IPv6 operation)
1567 When this value is changed from 1 to 0 (IPv6 is being enabled),
1568 it will dynamically create a link-local address on the given
1569 interface and start Duplicate Address Detection, if necessary.
1571 When this value is changed from 0 to 1 (IPv6 is being disabled),
1572 it will dynamically delete all address on the given interface.
1574 accept_dad - INTEGER
1575 Whether to accept DAD (Duplicate Address Detection).
1577 1: Enable DAD (default)
1578 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1579 link-local address has been found.
1581 force_tllao - BOOLEAN
1582 Enable sending the target link-layer address option even when
1583 responding to a unicast neighbor solicitation.
1586 Quoting from RFC 2461, section 4.4, Target link-layer address:
1588 "The option MUST be included for multicast solicitations in order to
1589 avoid infinite Neighbor Solicitation "recursion" when the peer node
1590 does not have a cache entry to return a Neighbor Advertisements
1591 message. When responding to unicast solicitations, the option can be
1592 omitted since the sender of the solicitation has the correct link-
1593 layer address; otherwise it would not have be able to send the unicast
1594 solicitation in the first place. However, including the link-layer
1595 address in this case adds little overhead and eliminates a potential
1596 race condition where the sender deletes the cached link-layer address
1597 prior to receiving a response to a previous solicitation."
1599 ndisc_notify - BOOLEAN
1600 Define mode for notification of address and device changes.
1601 0 - (default): do nothing
1602 1 - Generate unsolicited neighbour advertisements when device is brought
1603 up or hardware address changes.
1605 mldv1_unsolicited_report_interval - INTEGER
1606 The interval in milliseconds in which the next unsolicited
1607 MLDv1 report retransmit will take place.
1608 Default: 10000 (10 seconds)
1610 mldv2_unsolicited_report_interval - INTEGER
1611 The interval in milliseconds in which the next unsolicited
1612 MLDv2 report retransmit will take place.
1613 Default: 1000 (1 second)
1615 force_mld_version - INTEGER
1616 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1617 1 - Enforce to use MLD version 1
1618 2 - Enforce to use MLD version 2
1620 suppress_frag_ndisc - INTEGER
1621 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1622 with IPv6 Neighbor Discovery) behavior:
1623 1 - (default) discard fragmented neighbor discovery packets
1624 0 - allow fragmented neighbor discovery packets
1626 optimistic_dad - BOOLEAN
1627 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1628 0: disabled (default)
1631 use_optimistic - BOOLEAN
1632 If enabled, do not classify optimistic addresses as deprecated during
1633 source address selection. Preferred addresses will still be chosen
1634 before optimistic addresses, subject to other ranking in the source
1635 address selection algorithm.
1636 0: disabled (default)
1639 stable_secret - IPv6 address
1640 This IPv6 address will be used as a secret to generate IPv6
1641 addresses for link-local addresses and autoconfigured
1642 ones. All addresses generated after setting this secret will
1643 be stable privacy ones by default. This can be changed via the
1644 addrgenmode ip-link. conf/default/stable_secret is used as the
1645 secret for the namespace, the interface specific ones can
1646 overwrite that. Writes to conf/all/stable_secret are refused.
1648 It is recommended to generate this secret during installation
1649 of a system and keep it stable after that.
1651 By default the stable secret is unset.
1655 Limit the maximal rates for sending ICMPv6 packets.
1656 0 to disable any limiting,
1657 otherwise the minimal space between responses in milliseconds.
1660 xfrm6_gc_thresh - INTEGER
1661 The threshold at which we will start garbage collecting for IPv6
1662 destination cache entries. At twice this value the system will
1663 refuse new allocations.
1667 Pekka Savola <pekkas@netcore.fi>
1668 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1671 /proc/sys/net/bridge/* Variables:
1673 bridge-nf-call-arptables - BOOLEAN
1674 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1678 bridge-nf-call-iptables - BOOLEAN
1679 1 : pass bridged IPv4 traffic to iptables' chains.
1683 bridge-nf-call-ip6tables - BOOLEAN
1684 1 : pass bridged IPv6 traffic to ip6tables' chains.
1688 bridge-nf-filter-vlan-tagged - BOOLEAN
1689 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1693 bridge-nf-filter-pppoe-tagged - BOOLEAN
1694 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1698 bridge-nf-pass-vlan-input-dev - BOOLEAN
1699 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1700 interface on the bridge and set the netfilter input device to the vlan.
1701 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1702 target work with vlan-on-top-of-bridge interfaces. When no matching
1703 vlan interface is found, or this switch is off, the input device is
1704 set to the bridge interface.
1705 0: disable bridge netfilter vlan interface lookup.
1708 proc/sys/net/sctp/* Variables:
1710 addip_enable - BOOLEAN
1711 Enable or disable extension of Dynamic Address Reconfiguration
1712 (ADD-IP) functionality specified in RFC5061. This extension provides
1713 the ability to dynamically add and remove new addresses for the SCTP
1716 1: Enable extension.
1718 0: Disable extension.
1722 addip_noauth_enable - BOOLEAN
1723 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1724 authentication to protect the operations of adding or removing new
1725 addresses. This requirement is mandated so that unauthorized hosts
1726 would not be able to hijack associations. However, older
1727 implementations may not have implemented this requirement while
1728 allowing the ADD-IP extension. For reasons of interoperability,
1729 we provide this variable to control the enforcement of the
1730 authentication requirement.
1732 1: Allow ADD-IP extension to be used without authentication. This
1733 should only be set in a closed environment for interoperability
1734 with older implementations.
1736 0: Enforce the authentication requirement
1740 auth_enable - BOOLEAN
1741 Enable or disable Authenticated Chunks extension. This extension
1742 provides the ability to send and receive authenticated chunks and is
1743 required for secure operation of Dynamic Address Reconfiguration
1746 1: Enable this extension.
1747 0: Disable this extension.
1751 prsctp_enable - BOOLEAN
1752 Enable or disable the Partial Reliability extension (RFC3758) which
1753 is used to notify peers that a given DATA should no longer be expected.
1761 The limit of the number of new packets that can be initially sent. It
1762 controls how bursty the generated traffic can be.
1766 association_max_retrans - INTEGER
1767 Set the maximum number for retransmissions that an association can
1768 attempt deciding that the remote end is unreachable. If this value
1769 is exceeded, the association is terminated.
1773 max_init_retransmits - INTEGER
1774 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1775 that an association will attempt before declaring the destination
1776 unreachable and terminating.
1780 path_max_retrans - INTEGER
1781 The maximum number of retransmissions that will be attempted on a given
1782 path. Once this threshold is exceeded, the path is considered
1783 unreachable, and new traffic will use a different path when the
1784 association is multihomed.
1788 pf_retrans - INTEGER
1789 The number of retransmissions that will be attempted on a given path
1790 before traffic is redirected to an alternate transport (should one
1791 exist). Note this is distinct from path_max_retrans, as a path that
1792 passes the pf_retrans threshold can still be used. Its only
1793 deprioritized when a transmission path is selected by the stack. This
1794 setting is primarily used to enable fast failover mechanisms without
1795 having to reduce path_max_retrans to a very low value. See:
1796 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1797 for details. Note also that a value of pf_retrans > path_max_retrans
1798 disables this feature
1802 rto_initial - INTEGER
1803 The initial round trip timeout value in milliseconds that will be used
1804 in calculating round trip times. This is the initial time interval
1805 for retransmissions.
1810 The maximum value (in milliseconds) of the round trip timeout. This
1811 is the largest time interval that can elapse between retransmissions.
1816 The minimum value (in milliseconds) of the round trip timeout. This
1817 is the smallest time interval the can elapse between retransmissions.
1821 hb_interval - INTEGER
1822 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1823 are sent at the specified interval on idle paths to probe the state of
1824 a given path between 2 associations.
1828 sack_timeout - INTEGER
1829 The amount of time (in milliseconds) that the implementation will wait
1834 valid_cookie_life - INTEGER
1835 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1836 is used during association establishment.
1840 cookie_preserve_enable - BOOLEAN
1841 Enable or disable the ability to extend the lifetime of the SCTP cookie
1842 that is used during the establishment phase of SCTP association
1844 1: Enable cookie lifetime extension.
1849 cookie_hmac_alg - STRING
1850 Select the hmac algorithm used when generating the cookie value sent by
1851 a listening sctp socket to a connecting client in the INIT-ACK chunk.
1856 Ability to assign md5 or sha1 as the selected alg is predicated on the
1857 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
1858 CONFIG_CRYPTO_SHA1).
1860 Default: Dependent on configuration. MD5 if available, else SHA1 if
1861 available, else none.
1863 rcvbuf_policy - INTEGER
1864 Determines if the receive buffer is attributed to the socket or to
1865 association. SCTP supports the capability to create multiple
1866 associations on a single socket. When using this capability, it is
1867 possible that a single stalled association that's buffering a lot
1868 of data may block other associations from delivering their data by
1869 consuming all of the receive buffer space. To work around this,
1870 the rcvbuf_policy could be set to attribute the receiver buffer space
1871 to each association instead of the socket. This prevents the described
1874 1: rcvbuf space is per association
1875 0: rcvbuf space is per socket
1879 sndbuf_policy - INTEGER
1880 Similar to rcvbuf_policy above, this applies to send buffer space.
1882 1: Send buffer is tracked per association
1883 0: Send buffer is tracked per socket.
1887 sctp_mem - vector of 3 INTEGERs: min, pressure, max
1888 Number of pages allowed for queueing by all SCTP sockets.
1890 min: Below this number of pages SCTP is not bothered about its
1891 memory appetite. When amount of memory allocated by SCTP exceeds
1892 this number, SCTP starts to moderate memory usage.
1894 pressure: This value was introduced to follow format of tcp_mem.
1896 max: Number of pages allowed for queueing by all SCTP sockets.
1898 Default is calculated at boot time from amount of available memory.
1900 sctp_rmem - vector of 3 INTEGERs: min, default, max
1901 Only the first value ("min") is used, "default" and "max" are
1904 min: Minimal size of receive buffer used by SCTP socket.
1905 It is guaranteed to each SCTP socket (but not association) even
1906 under moderate memory pressure.
1910 sctp_wmem - vector of 3 INTEGERs: min, default, max
1911 Currently this tunable has no effect.
1913 addr_scope_policy - INTEGER
1914 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1916 0 - Disable IPv4 address scoping
1917 1 - Enable IPv4 address scoping
1918 2 - Follow draft but allow IPv4 private addresses
1919 3 - Follow draft but allow IPv4 link local addresses
1924 /proc/sys/net/core/*
1925 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
1928 /proc/sys/net/unix/*
1929 max_dgram_qlen - INTEGER
1930 The maximum length of dgram socket receive queue
1937 /proc/sys/net/irda/*
1938 fast_poll_increase FIXME
1939 warn_noreply_time FIXME
1940 discovery_slots FIXME
1943 discovery_timeout FIXME
1944 lap_keepalive_time FIXME
1945 max_noreply_time FIXME
1946 max_tx_data_size FIXME
1948 min_tx_turn_time FIXME