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 fib_multipath_use_neigh - BOOLEAN
67 Use status of existing neighbor entry when determining nexthop for
68 multipath routes. If disabled, neighbor information is not used and
69 packets could be directed to a failed nexthop. Only valid for kernels
70 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
76 route/max_size - INTEGER
77 Maximum number of routes allowed in the kernel. Increase
78 this when using large numbers of interfaces and/or routes.
79 From linux kernel 3.6 onwards, this is deprecated for ipv4
80 as route cache is no longer used.
82 neigh/default/gc_thresh1 - INTEGER
83 Minimum number of entries to keep. Garbage collector will not
84 purge entries if there are fewer than this number.
87 neigh/default/gc_thresh2 - INTEGER
88 Threshold when garbage collector becomes more aggressive about
89 purging entries. Entries older than 5 seconds will be cleared
90 when over this number.
93 neigh/default/gc_thresh3 - INTEGER
94 Maximum number of neighbor entries allowed. Increase this
95 when using large numbers of interfaces and when communicating
96 with large numbers of directly-connected peers.
99 neigh/default/unres_qlen_bytes - INTEGER
100 The maximum number of bytes which may be used by packets
101 queued for each unresolved address by other network layers.
103 Setting negative value is meaningless and will return error.
104 Default: 65536 Bytes(64KB)
106 neigh/default/unres_qlen - INTEGER
107 The maximum number of packets which may be queued for each
108 unresolved address by other network layers.
109 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
110 Prior to linux 3.3, the default value is 3 which may cause
111 unexpected packet loss. The current default value is calculated
112 according to default value of unres_qlen_bytes and true size of
116 mtu_expires - INTEGER
117 Time, in seconds, that cached PMTU information is kept.
119 min_adv_mss - INTEGER
120 The advertised MSS depends on the first hop route MTU, but will
121 never be lower than this setting.
125 ipfrag_high_thresh - INTEGER
126 Maximum memory used to reassemble IP fragments. When
127 ipfrag_high_thresh bytes of memory is allocated for this purpose,
128 the fragment handler will toss packets until ipfrag_low_thresh
129 is reached. This also serves as a maximum limit to namespaces
130 different from the initial one.
132 ipfrag_low_thresh - INTEGER
133 Maximum memory used to reassemble IP fragments before the kernel
134 begins to remove incomplete fragment queues to free up resources.
135 The kernel still accepts new fragments for defragmentation.
137 ipfrag_time - INTEGER
138 Time in seconds to keep an IP fragment in memory.
140 ipfrag_max_dist - INTEGER
141 ipfrag_max_dist is a non-negative integer value which defines the
142 maximum "disorder" which is allowed among fragments which share a
143 common IP source address. Note that reordering of packets is
144 not unusual, but if a large number of fragments arrive from a source
145 IP address while a particular fragment queue remains incomplete, it
146 probably indicates that one or more fragments belonging to that queue
147 have been lost. When ipfrag_max_dist is positive, an additional check
148 is done on fragments before they are added to a reassembly queue - if
149 ipfrag_max_dist (or more) fragments have arrived from a particular IP
150 address between additions to any IP fragment queue using that source
151 address, it's presumed that one or more fragments in the queue are
152 lost. The existing fragment queue will be dropped, and a new one
153 started. An ipfrag_max_dist value of zero disables this check.
155 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
156 result in unnecessarily dropping fragment queues when normal
157 reordering of packets occurs, which could lead to poor application
158 performance. Using a very large value, e.g. 50000, increases the
159 likelihood of incorrectly reassembling IP fragments that originate
160 from different IP datagrams, which could result in data corruption.
165 inet_peer_threshold - INTEGER
166 The approximate size of the storage. Starting from this threshold
167 entries will be thrown aggressively. This threshold also determines
168 entries' time-to-live and time intervals between garbage collection
169 passes. More entries, less time-to-live, less GC interval.
171 inet_peer_minttl - INTEGER
172 Minimum time-to-live of entries. Should be enough to cover fragment
173 time-to-live on the reassembling side. This minimum time-to-live is
174 guaranteed if the pool size is less than inet_peer_threshold.
177 inet_peer_maxttl - INTEGER
178 Maximum time-to-live of entries. Unused entries will expire after
179 this period of time if there is no memory pressure on the pool (i.e.
180 when the number of entries in the pool is very small).
186 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
187 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
190 tcp_abort_on_overflow - BOOLEAN
191 If listening service is too slow to accept new connections,
192 reset them. Default state is FALSE. It means that if overflow
193 occurred due to a burst, connection will recover. Enable this
194 option _only_ if you are really sure that listening daemon
195 cannot be tuned to accept connections faster. Enabling this
196 option can harm clients of your server.
198 tcp_adv_win_scale - INTEGER
199 Count buffering overhead as bytes/2^tcp_adv_win_scale
200 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
202 Possible values are [-31, 31], inclusive.
205 tcp_allowed_congestion_control - STRING
206 Show/set the congestion control choices available to non-privileged
207 processes. The list is a subset of those listed in
208 tcp_available_congestion_control.
209 Default is "reno" and the default setting (tcp_congestion_control).
211 tcp_app_win - INTEGER
212 Reserve max(window/2^tcp_app_win, mss) of window for application
213 buffer. Value 0 is special, it means that nothing is reserved.
216 tcp_autocorking - BOOLEAN
217 Enable TCP auto corking :
218 When applications do consecutive small write()/sendmsg() system calls,
219 we try to coalesce these small writes as much as possible, to lower
220 total amount of sent packets. This is done if at least one prior
221 packet for the flow is waiting in Qdisc queues or device transmit
222 queue. Applications can still use TCP_CORK for optimal behavior
223 when they know how/when to uncork their sockets.
226 tcp_available_congestion_control - STRING
227 Shows the available congestion control choices that are registered.
228 More congestion control algorithms may be available as modules,
231 tcp_base_mss - INTEGER
232 The initial value of search_low to be used by the packetization layer
233 Path MTU discovery (MTU probing). If MTU probing is enabled,
234 this is the initial MSS used by the connection.
236 tcp_congestion_control - STRING
237 Set the congestion control algorithm to be used for new
238 connections. The algorithm "reno" is always available, but
239 additional choices may be available based on kernel configuration.
240 Default is set as part of kernel configuration.
241 For passive connections, the listener congestion control choice
243 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
246 Allows TCP to send "duplicate" SACKs.
248 tcp_early_retrans - INTEGER
249 Tail loss probe (TLP) converts RTOs occurring due to tail
250 losses into fast recovery (draft-ietf-tcpm-rack). Note that
251 TLP requires RACK to function properly (see tcp_recovery below)
258 Control use of Explicit Congestion Notification (ECN) by TCP.
259 ECN is used only when both ends of the TCP connection indicate
260 support for it. This feature is useful in avoiding losses due
261 to congestion by allowing supporting routers to signal
262 congestion before having to drop packets.
264 0 Disable ECN. Neither initiate nor accept ECN.
265 1 Enable ECN when requested by incoming connections and
266 also request ECN on outgoing connection attempts.
267 2 Enable ECN when requested by incoming connections
268 but do not request ECN on outgoing connections.
271 tcp_ecn_fallback - BOOLEAN
272 If the kernel detects that ECN connection misbehaves, enable fall
273 back to non-ECN. Currently, this knob implements the fallback
274 from RFC3168, section 6.1.1.1., but we reserve that in future,
275 additional detection mechanisms could be implemented under this
276 knob. The value is not used, if tcp_ecn or per route (or congestion
277 control) ECN settings are disabled.
278 Default: 1 (fallback enabled)
281 Enable FACK congestion avoidance and fast retransmission.
282 The value is not used, if tcp_sack is not enabled.
284 tcp_fin_timeout - INTEGER
285 The length of time an orphaned (no longer referenced by any
286 application) connection will remain in the FIN_WAIT_2 state
287 before it is aborted at the local end. While a perfectly
288 valid "receive only" state for an un-orphaned connection, an
289 orphaned connection in FIN_WAIT_2 state could otherwise wait
290 forever for the remote to close its end of the connection.
295 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
296 F-RTO is an enhanced recovery algorithm for TCP retransmission
297 timeouts. It is particularly beneficial in networks where the
298 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
299 modification. It does not require any support from the peer.
301 By default it's enabled with a non-zero value. 0 disables F-RTO.
303 tcp_invalid_ratelimit - INTEGER
304 Limit the maximal rate for sending duplicate acknowledgments
305 in response to incoming TCP packets that are for an existing
306 connection but that are invalid due to any of these reasons:
308 (a) out-of-window sequence number,
309 (b) out-of-window acknowledgment number, or
310 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
312 This can help mitigate simple "ack loop" DoS attacks, wherein
313 a buggy or malicious middlebox or man-in-the-middle can
314 rewrite TCP header fields in manner that causes each endpoint
315 to think that the other is sending invalid TCP segments, thus
316 causing each side to send an unterminating stream of duplicate
317 acknowledgments for invalid segments.
319 Using 0 disables rate-limiting of dupacks in response to
320 invalid segments; otherwise this value specifies the minimal
321 space between sending such dupacks, in milliseconds.
323 Default: 500 (milliseconds).
325 tcp_keepalive_time - INTEGER
326 How often TCP sends out keepalive messages when keepalive is enabled.
329 tcp_keepalive_probes - INTEGER
330 How many keepalive probes TCP sends out, until it decides that the
331 connection is broken. Default value: 9.
333 tcp_keepalive_intvl - INTEGER
334 How frequently the probes are send out. Multiplied by
335 tcp_keepalive_probes it is time to kill not responding connection,
336 after probes started. Default value: 75sec i.e. connection
337 will be aborted after ~11 minutes of retries.
339 tcp_l3mdev_accept - BOOLEAN
340 Enables child sockets to inherit the L3 master device index.
341 Enabling this option allows a "global" listen socket to work
342 across L3 master domains (e.g., VRFs) with connected sockets
343 derived from the listen socket to be bound to the L3 domain in
344 which the packets originated. Only valid when the kernel was
345 compiled with CONFIG_NET_L3_MASTER_DEV.
347 tcp_low_latency - BOOLEAN
348 If set, the TCP stack makes decisions that prefer lower
349 latency as opposed to higher throughput. By default, this
350 option is not set meaning that higher throughput is preferred.
351 An example of an application where this default should be
352 changed would be a Beowulf compute cluster.
355 tcp_max_orphans - INTEGER
356 Maximal number of TCP sockets not attached to any user file handle,
357 held by system. If this number is exceeded orphaned connections are
358 reset immediately and warning is printed. This limit exists
359 only to prevent simple DoS attacks, you _must_ not rely on this
360 or lower the limit artificially, but rather increase it
361 (probably, after increasing installed memory),
362 if network conditions require more than default value,
363 and tune network services to linger and kill such states
364 more aggressively. Let me to remind again: each orphan eats
365 up to ~64K of unswappable memory.
367 tcp_max_syn_backlog - INTEGER
368 Maximal number of remembered connection requests, which have not
369 received an acknowledgment from connecting client.
370 The minimal value is 128 for low memory machines, and it will
371 increase in proportion to the memory of machine.
372 If server suffers from overload, try increasing this number.
374 tcp_max_tw_buckets - INTEGER
375 Maximal number of timewait sockets held by system simultaneously.
376 If this number is exceeded time-wait socket is immediately destroyed
377 and warning is printed. This limit exists only to prevent
378 simple DoS attacks, you _must_ not lower the limit artificially,
379 but rather increase it (probably, after increasing installed memory),
380 if network conditions require more than default value.
382 tcp_mem - vector of 3 INTEGERs: min, pressure, max
383 min: below this number of pages TCP is not bothered about its
386 pressure: when amount of memory allocated by TCP exceeds this number
387 of pages, TCP moderates its memory consumption and enters memory
388 pressure mode, which is exited when memory consumption falls
391 max: number of pages allowed for queueing by all TCP sockets.
393 Defaults are calculated at boot time from amount of available
396 tcp_min_rtt_wlen - INTEGER
397 The window length of the windowed min filter to track the minimum RTT.
398 A shorter window lets a flow more quickly pick up new (higher)
399 minimum RTT when it is moved to a longer path (e.g., due to traffic
400 engineering). A longer window makes the filter more resistant to RTT
401 inflations such as transient congestion. The unit is seconds.
404 tcp_moderate_rcvbuf - BOOLEAN
405 If set, TCP performs receive buffer auto-tuning, attempting to
406 automatically size the buffer (no greater than tcp_rmem[2]) to
407 match the size required by the path for full throughput. Enabled by
410 tcp_mtu_probing - INTEGER
411 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
414 1 - Disabled by default, enabled when an ICMP black hole detected
415 2 - Always enabled, use initial MSS of tcp_base_mss.
417 tcp_probe_interval - INTEGER
418 Controls how often to start TCP Packetization-Layer Path MTU
419 Discovery reprobe. The default is reprobing every 10 minutes as
422 tcp_probe_threshold - INTEGER
423 Controls when TCP Packetization-Layer Path MTU Discovery probing
424 will stop in respect to the width of search range in bytes. Default
427 tcp_no_metrics_save - BOOLEAN
428 By default, TCP saves various connection metrics in the route cache
429 when the connection closes, so that connections established in the
430 near future can use these to set initial conditions. Usually, this
431 increases overall performance, but may sometimes cause performance
432 degradation. If set, TCP will not cache metrics on closing
435 tcp_orphan_retries - INTEGER
436 This value influences the timeout of a locally closed TCP connection,
437 when RTO retransmissions remain unacknowledged.
438 See tcp_retries2 for more details.
440 The default value is 8.
441 If your machine is a loaded WEB server,
442 you should think about lowering this value, such sockets
443 may consume significant resources. Cf. tcp_max_orphans.
445 tcp_recovery - INTEGER
446 This value is a bitmap to enable various experimental loss recovery
449 RACK: 0x1 enables the RACK loss detection for fast detection of lost
450 retransmissions and tail drops.
454 tcp_reordering - INTEGER
455 Initial reordering level of packets in a TCP stream.
456 TCP stack can then dynamically adjust flow reordering level
457 between this initial value and tcp_max_reordering
460 tcp_max_reordering - INTEGER
461 Maximal reordering level of packets in a TCP stream.
462 300 is a fairly conservative value, but you might increase it
463 if paths are using per packet load balancing (like bonding rr mode)
466 tcp_retrans_collapse - BOOLEAN
467 Bug-to-bug compatibility with some broken printers.
468 On retransmit try to send bigger packets to work around bugs in
471 tcp_retries1 - INTEGER
472 This value influences the time, after which TCP decides, that
473 something is wrong due to unacknowledged RTO retransmissions,
474 and reports this suspicion to the network layer.
475 See tcp_retries2 for more details.
477 RFC 1122 recommends at least 3 retransmissions, which is the
480 tcp_retries2 - INTEGER
481 This value influences the timeout of an alive TCP connection,
482 when RTO retransmissions remain unacknowledged.
483 Given a value of N, a hypothetical TCP connection following
484 exponential backoff with an initial RTO of TCP_RTO_MIN would
485 retransmit N times before killing the connection at the (N+1)th RTO.
487 The default value of 15 yields a hypothetical timeout of 924.6
488 seconds and is a lower bound for the effective timeout.
489 TCP will effectively time out at the first RTO which exceeds the
490 hypothetical timeout.
492 RFC 1122 recommends at least 100 seconds for the timeout,
493 which corresponds to a value of at least 8.
495 tcp_rfc1337 - BOOLEAN
496 If set, the TCP stack behaves conforming to RFC1337. If unset,
497 we are not conforming to RFC, but prevent TCP TIME_WAIT
501 tcp_rmem - vector of 3 INTEGERs: min, default, max
502 min: Minimal size of receive buffer used by TCP sockets.
503 It is guaranteed to each TCP socket, even under moderate memory
507 default: initial size of receive buffer used by TCP sockets.
508 This value overrides net.core.rmem_default used by other protocols.
509 Default: 87380 bytes. This value results in window of 65535 with
510 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
511 less for default tcp_app_win. See below about these variables.
513 max: maximal size of receive buffer allowed for automatically
514 selected receiver buffers for TCP socket. This value does not override
515 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
516 automatic tuning of that socket's receive buffer size, in which
517 case this value is ignored.
518 Default: between 87380B and 6MB, depending on RAM size.
521 Enable select acknowledgments (SACKS).
523 tcp_slow_start_after_idle - BOOLEAN
524 If set, provide RFC2861 behavior and time out the congestion
525 window after an idle period. An idle period is defined at
526 the current RTO. If unset, the congestion window will not
527 be timed out after an idle period.
531 Use the Host requirements interpretation of the TCP urgent pointer field.
532 Most hosts use the older BSD interpretation, so if you turn this on
533 Linux might not communicate correctly with them.
536 tcp_synack_retries - INTEGER
537 Number of times SYNACKs for a passive TCP connection attempt will
538 be retransmitted. Should not be higher than 255. Default value
539 is 5, which corresponds to 31seconds till the last retransmission
540 with the current initial RTO of 1second. With this the final timeout
541 for a passive TCP connection will happen after 63seconds.
543 tcp_syncookies - BOOLEAN
544 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
545 Send out syncookies when the syn backlog queue of a socket
546 overflows. This is to prevent against the common 'SYN flood attack'
549 Note, that syncookies is fallback facility.
550 It MUST NOT be used to help highly loaded servers to stand
551 against legal connection rate. If you see SYN flood warnings
552 in your logs, but investigation shows that they occur
553 because of overload with legal connections, you should tune
554 another parameters until this warning disappear.
555 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
557 syncookies seriously violate TCP protocol, do not allow
558 to use TCP extensions, can result in serious degradation
559 of some services (f.e. SMTP relaying), visible not by you,
560 but your clients and relays, contacting you. While you see
561 SYN flood warnings in logs not being really flooded, your server
562 is seriously misconfigured.
564 If you want to test which effects syncookies have to your
565 network connections you can set this knob to 2 to enable
566 unconditionally generation of syncookies.
568 tcp_fastopen - INTEGER
569 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
572 The client support is enabled by flag 0x1 (on by default). The client
573 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
574 rather than connect() to send data in SYN.
576 The server support is enabled by flag 0x2 (off by default). Then
577 either enable for all listeners with another flag (0x400) or
578 enable individual listeners via TCP_FASTOPEN socket option with
579 the option value being the length of the syn-data backlog.
581 The values (bitmap) are
582 0x1: (client) enables sending data in the opening SYN on the client.
583 0x2: (server) enables the server support, i.e., allowing data in
584 a SYN packet to be accepted and passed to the
585 application before 3-way handshake finishes.
586 0x4: (client) send data in the opening SYN regardless of cookie
587 availability and without a cookie option.
588 0x200: (server) accept data-in-SYN w/o any cookie option present.
589 0x400: (server) enable all listeners to support Fast Open by
590 default without explicit TCP_FASTOPEN socket option.
594 Note that that additional client or server features are only
595 effective if the basic support (0x1 and 0x2) are enabled respectively.
597 tcp_syn_retries - INTEGER
598 Number of times initial SYNs for an active TCP connection attempt
599 will be retransmitted. Should not be higher than 127. Default value
600 is 6, which corresponds to 63seconds till the last retransmission
601 with the current initial RTO of 1second. With this the final timeout
602 for an active TCP connection attempt will happen after 127seconds.
604 tcp_timestamps - INTEGER
605 Enable timestamps as defined in RFC1323.
607 1: Enable timestamps as defined in RFC1323 and use random offset for
608 each connection rather than only using the current time.
609 2: Like 1, but without random offsets.
612 tcp_min_tso_segs - INTEGER
613 Minimal number of segments per TSO frame.
614 Since linux-3.12, TCP does an automatic sizing of TSO frames,
615 depending on flow rate, instead of filling 64Kbytes packets.
616 For specific usages, it's possible to force TCP to build big
617 TSO frames. Note that TCP stack might split too big TSO packets
618 if available window is too small.
621 tcp_pacing_ss_ratio - INTEGER
622 sk->sk_pacing_rate is set by TCP stack using a ratio applied
623 to current rate. (current_rate = cwnd * mss / srtt)
624 If TCP is in slow start, tcp_pacing_ss_ratio is applied
625 to let TCP probe for bigger speeds, assuming cwnd can be
626 doubled every other RTT.
629 tcp_pacing_ca_ratio - INTEGER
630 sk->sk_pacing_rate is set by TCP stack using a ratio applied
631 to current rate. (current_rate = cwnd * mss / srtt)
632 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
633 is applied to conservatively probe for bigger throughput.
636 tcp_tso_win_divisor - INTEGER
637 This allows control over what percentage of the congestion window
638 can be consumed by a single TSO frame.
639 The setting of this parameter is a choice between burstiness and
640 building larger TSO frames.
643 tcp_tw_recycle - BOOLEAN
644 Enable fast recycling TIME-WAIT sockets. Default value is 0.
645 It should not be changed without advice/request of technical
648 tcp_tw_reuse - BOOLEAN
649 Allow to reuse TIME-WAIT sockets for new connections when it is
650 safe from protocol viewpoint. Default value is 0.
651 It should not be changed without advice/request of technical
654 tcp_window_scaling - BOOLEAN
655 Enable window scaling as defined in RFC1323.
657 tcp_wmem - vector of 3 INTEGERs: min, default, max
658 min: Amount of memory reserved for send buffers for TCP sockets.
659 Each TCP socket has rights to use it due to fact of its birth.
662 default: initial size of send buffer used by TCP sockets. This
663 value overrides net.core.wmem_default used by other protocols.
664 It is usually lower than net.core.wmem_default.
667 max: Maximal amount of memory allowed for automatically tuned
668 send buffers for TCP sockets. This value does not override
669 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
670 automatic tuning of that socket's send buffer size, in which case
671 this value is ignored.
672 Default: between 64K and 4MB, depending on RAM size.
674 tcp_notsent_lowat - UNSIGNED INTEGER
675 A TCP socket can control the amount of unsent bytes in its write queue,
676 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
677 reports POLLOUT events if the amount of unsent bytes is below a per
678 socket value, and if the write queue is not full. sendmsg() will
679 also not add new buffers if the limit is hit.
681 This global variable controls the amount of unsent data for
682 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
683 to the global variable has immediate effect.
685 Default: UINT_MAX (0xFFFFFFFF)
687 tcp_workaround_signed_windows - BOOLEAN
688 If set, assume no receipt of a window scaling option means the
689 remote TCP is broken and treats the window as a signed quantity.
690 If unset, assume the remote TCP is not broken even if we do
691 not receive a window scaling option from them.
694 tcp_thin_linear_timeouts - BOOLEAN
695 Enable dynamic triggering of linear timeouts for thin streams.
696 If set, a check is performed upon retransmission by timeout to
697 determine if the stream is thin (less than 4 packets in flight).
698 As long as the stream is found to be thin, up to 6 linear
699 timeouts may be performed before exponential backoff mode is
700 initiated. This improves retransmission latency for
701 non-aggressive thin streams, often found to be time-dependent.
702 For more information on thin streams, see
703 Documentation/networking/tcp-thin.txt
706 tcp_limit_output_bytes - INTEGER
707 Controls TCP Small Queue limit per tcp socket.
708 TCP bulk sender tends to increase packets in flight until it
709 gets losses notifications. With SNDBUF autotuning, this can
710 result in a large amount of packets queued in qdisc/device
711 on the local machine, hurting latency of other flows, for
712 typical pfifo_fast qdiscs.
713 tcp_limit_output_bytes limits the number of bytes on qdisc
714 or device to reduce artificial RTT/cwnd and reduce bufferbloat.
717 tcp_challenge_ack_limit - INTEGER
718 Limits number of Challenge ACK sent per second, as recommended
719 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
724 udp_mem - vector of 3 INTEGERs: min, pressure, max
725 Number of pages allowed for queueing by all UDP sockets.
727 min: Below this number of pages UDP is not bothered about its
728 memory appetite. When amount of memory allocated by UDP exceeds
729 this number, UDP starts to moderate memory usage.
731 pressure: This value was introduced to follow format of tcp_mem.
733 max: Number of pages allowed for queueing by all UDP sockets.
735 Default is calculated at boot time from amount of available memory.
737 udp_rmem_min - INTEGER
738 Minimal size of receive buffer used by UDP sockets in moderation.
739 Each UDP socket is able to use the size for receiving data, even if
740 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
743 udp_wmem_min - INTEGER
744 Minimal size of send buffer used by UDP sockets in moderation.
745 Each UDP socket is able to use the size for sending data, even if
746 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
751 cipso_cache_enable - BOOLEAN
752 If set, enable additions to and lookups from the CIPSO label mapping
753 cache. If unset, additions are ignored and lookups always result in a
754 miss. However, regardless of the setting the cache is still
755 invalidated when required when means you can safely toggle this on and
756 off and the cache will always be "safe".
759 cipso_cache_bucket_size - INTEGER
760 The CIPSO label cache consists of a fixed size hash table with each
761 hash bucket containing a number of cache entries. This variable limits
762 the number of entries in each hash bucket; the larger the value the
763 more CIPSO label mappings that can be cached. When the number of
764 entries in a given hash bucket reaches this limit adding new entries
765 causes the oldest entry in the bucket to be removed to make room.
768 cipso_rbm_optfmt - BOOLEAN
769 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
770 the CIPSO draft specification (see Documentation/netlabel for details).
771 This means that when set the CIPSO tag will be padded with empty
772 categories in order to make the packet data 32-bit aligned.
775 cipso_rbm_structvalid - BOOLEAN
776 If set, do a very strict check of the CIPSO option when
777 ip_options_compile() is called. If unset, relax the checks done during
778 ip_options_compile(). Either way is "safe" as errors are caught else
779 where in the CIPSO processing code but setting this to 0 (False) should
780 result in less work (i.e. it should be faster) but could cause problems
781 with other implementations that require strict checking.
786 ip_local_port_range - 2 INTEGERS
787 Defines the local port range that is used by TCP and UDP to
788 choose the local port. The first number is the first, the
789 second the last local port number.
790 If possible, it is better these numbers have different parity.
791 (one even and one odd values)
792 The default values are 32768 and 60999 respectively.
794 ip_local_reserved_ports - list of comma separated ranges
795 Specify the ports which are reserved for known third-party
796 applications. These ports will not be used by automatic port
797 assignments (e.g. when calling connect() or bind() with port
798 number 0). Explicit port allocation behavior is unchanged.
800 The format used for both input and output is a comma separated
801 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
802 10). Writing to the file will clear all previously reserved
803 ports and update the current list with the one given in the
806 Note that ip_local_port_range and ip_local_reserved_ports
807 settings are independent and both are considered by the kernel
808 when determining which ports are available for automatic port
811 You can reserve ports which are not in the current
812 ip_local_port_range, e.g.:
814 $ cat /proc/sys/net/ipv4/ip_local_port_range
816 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
819 although this is redundant. However such a setting is useful
820 if later the port range is changed to a value that will
821 include the reserved ports.
825 ip_unprivileged_port_start - INTEGER
826 This is a per-namespace sysctl. It defines the first
827 unprivileged port in the network namespace. Privileged ports
828 require root or CAP_NET_BIND_SERVICE in order to bind to them.
829 To disable all privileged ports, set this to 0. It may not
830 overlap with the ip_local_reserved_ports range.
834 ip_nonlocal_bind - BOOLEAN
835 If set, allows processes to bind() to non-local IP addresses,
836 which can be quite useful - but may break some applications.
840 If set non-zero, enables support for dynamic addresses.
841 If set to a non-zero value larger than 1, a kernel log
842 message will be printed when dynamic address rewriting
846 ip_early_demux - BOOLEAN
847 Optimize input packet processing down to one demux for
848 certain kinds of local sockets. Currently we only do this
849 for established TCP sockets.
851 It may add an additional cost for pure routing workloads that
852 reduces overall throughput, in such case you should disable it.
855 icmp_echo_ignore_all - BOOLEAN
856 If set non-zero, then the kernel will ignore all ICMP ECHO
860 icmp_echo_ignore_broadcasts - BOOLEAN
861 If set non-zero, then the kernel will ignore all ICMP ECHO and
862 TIMESTAMP requests sent to it via broadcast/multicast.
865 icmp_ratelimit - INTEGER
866 Limit the maximal rates for sending ICMP packets whose type matches
867 icmp_ratemask (see below) to specific targets.
868 0 to disable any limiting,
869 otherwise the minimal space between responses in milliseconds.
870 Note that another sysctl, icmp_msgs_per_sec limits the number
871 of ICMP packets sent on all targets.
874 icmp_msgs_per_sec - INTEGER
875 Limit maximal number of ICMP packets sent per second from this host.
876 Only messages whose type matches icmp_ratemask (see below) are
877 controlled by this limit.
880 icmp_msgs_burst - INTEGER
881 icmp_msgs_per_sec controls number of ICMP packets sent per second,
882 while icmp_msgs_burst controls the burst size of these packets.
885 icmp_ratemask - INTEGER
886 Mask made of ICMP types for which rates are being limited.
887 Significant bits: IHGFEDCBA9876543210
888 Default mask: 0000001100000011000 (6168)
890 Bit definitions (see include/linux/icmp.h):
892 3 Destination Unreachable *
897 C Parameter Problem *
902 H Address Mask Request
905 * These are rate limited by default (see default mask above)
907 icmp_ignore_bogus_error_responses - BOOLEAN
908 Some routers violate RFC1122 by sending bogus responses to broadcast
909 frames. Such violations are normally logged via a kernel warning.
910 If this is set to TRUE, the kernel will not give such warnings, which
911 will avoid log file clutter.
914 icmp_errors_use_inbound_ifaddr - BOOLEAN
916 If zero, icmp error messages are sent with the primary address of
917 the exiting interface.
919 If non-zero, the message will be sent with the primary address of
920 the interface that received the packet that caused the icmp error.
921 This is the behaviour network many administrators will expect from
922 a router. And it can make debugging complicated network layouts
925 Note that if no primary address exists for the interface selected,
926 then the primary address of the first non-loopback interface that
927 has one will be used regardless of this setting.
931 igmp_max_memberships - INTEGER
932 Change the maximum number of multicast groups we can subscribe to.
935 Theoretical maximum value is bounded by having to send a membership
936 report in a single datagram (i.e. the report can't span multiple
937 datagrams, or risk confusing the switch and leaving groups you don't
940 The number of supported groups 'M' is bounded by the number of group
941 report entries you can fit into a single datagram of 65535 bytes.
943 M = 65536-sizeof (ip header)/(sizeof(Group record))
945 Group records are variable length, with a minimum of 12 bytes.
946 So net.ipv4.igmp_max_memberships should not be set higher than:
948 (65536-24) / 12 = 5459
950 The value 5459 assumes no IP header options, so in practice
951 this number may be lower.
953 igmp_max_msf - INTEGER
954 Maximum number of addresses allowed in the source filter list for a
959 Controls the IGMP query robustness variable (see RFC2236 8.1).
960 Default: 2 (as specified by RFC2236 8.1)
961 Minimum: 1 (as specified by RFC6636 4.5)
963 force_igmp_version - INTEGER
964 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
965 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
966 Present timer expires.
967 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
968 receive IGMPv2/v3 query.
969 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
970 IGMPv1 query message. Will reply report if receive IGMPv3 query.
971 3 - Enforce to use IGMP version 3. The same react with default 0.
973 Note: this is not the same with force_mld_version because IGMPv3 RFC3376
974 Security Considerations does not have clear description that we could
975 ignore other version messages completely as MLDv2 RFC3810. So make
976 this value as default 0 is recommended.
978 conf/interface/* changes special settings per interface (where
979 "interface" is the name of your network interface)
981 conf/all/* is special, changes the settings for all interfaces
983 log_martians - BOOLEAN
984 Log packets with impossible addresses to kernel log.
985 log_martians for the interface will be enabled if at least one of
986 conf/{all,interface}/log_martians is set to TRUE,
987 it will be disabled otherwise
989 accept_redirects - BOOLEAN
990 Accept ICMP redirect messages.
991 accept_redirects for the interface will be enabled if:
992 - both conf/{all,interface}/accept_redirects are TRUE in the case
993 forwarding for the interface is enabled
995 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
996 case forwarding for the interface is disabled
997 accept_redirects for the interface will be disabled otherwise
1001 forwarding - BOOLEAN
1002 Enable IP forwarding on this interface.
1004 mc_forwarding - BOOLEAN
1005 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1006 and a multicast routing daemon is required.
1007 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1008 routing for the interface
1011 Integer value used to differentiate the devices by the medium they
1012 are attached to. Two devices can have different id values when
1013 the broadcast packets are received only on one of them.
1014 The default value 0 means that the device is the only interface
1015 to its medium, value of -1 means that medium is not known.
1017 Currently, it is used to change the proxy_arp behavior:
1018 the proxy_arp feature is enabled for packets forwarded between
1019 two devices attached to different media.
1023 proxy_arp for the interface will be enabled if at least one of
1024 conf/{all,interface}/proxy_arp is set to TRUE,
1025 it will be disabled otherwise
1027 proxy_arp_pvlan - BOOLEAN
1028 Private VLAN proxy arp.
1029 Basically allow proxy arp replies back to the same interface
1030 (from which the ARP request/solicitation was received).
1032 This is done to support (ethernet) switch features, like RFC
1033 3069, where the individual ports are NOT allowed to
1034 communicate with each other, but they are allowed to talk to
1035 the upstream router. As described in RFC 3069, it is possible
1036 to allow these hosts to communicate through the upstream
1037 router by proxy_arp'ing. Don't need to be used together with
1040 This technology is known by different names:
1041 In RFC 3069 it is called VLAN Aggregation.
1042 Cisco and Allied Telesyn call it Private VLAN.
1043 Hewlett-Packard call it Source-Port filtering or port-isolation.
1044 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1046 shared_media - BOOLEAN
1047 Send(router) or accept(host) RFC1620 shared media redirects.
1048 Overrides secure_redirects.
1049 shared_media for the interface will be enabled if at least one of
1050 conf/{all,interface}/shared_media is set to TRUE,
1051 it will be disabled otherwise
1054 secure_redirects - BOOLEAN
1055 Accept ICMP redirect messages only to gateways listed in the
1056 interface's current gateway list. Even if disabled, RFC1122 redirect
1058 Overridden by shared_media.
1059 secure_redirects for the interface will be enabled if at least one of
1060 conf/{all,interface}/secure_redirects is set to TRUE,
1061 it will be disabled otherwise
1064 send_redirects - BOOLEAN
1065 Send redirects, if router.
1066 send_redirects for the interface will be enabled if at least one of
1067 conf/{all,interface}/send_redirects is set to TRUE,
1068 it will be disabled otherwise
1071 bootp_relay - BOOLEAN
1072 Accept packets with source address 0.b.c.d destined
1073 not to this host as local ones. It is supposed, that
1074 BOOTP relay daemon will catch and forward such packets.
1075 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1078 Not Implemented Yet.
1080 accept_source_route - BOOLEAN
1081 Accept packets with SRR option.
1082 conf/all/accept_source_route must also be set to TRUE to accept packets
1083 with SRR option on the interface
1084 default TRUE (router)
1087 accept_local - BOOLEAN
1088 Accept packets with local source addresses. In combination with
1089 suitable routing, this can be used to direct packets between two
1090 local interfaces over the wire and have them accepted properly.
1093 route_localnet - BOOLEAN
1094 Do not consider loopback addresses as martian source or destination
1095 while routing. This enables the use of 127/8 for local routing purposes.
1099 0 - No source validation.
1100 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1101 Each incoming packet is tested against the FIB and if the interface
1102 is not the best reverse path the packet check will fail.
1103 By default failed packets are discarded.
1104 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1105 Each incoming packet's source address is also tested against the FIB
1106 and if the source address is not reachable via any interface
1107 the packet check will fail.
1109 Current recommended practice in RFC3704 is to enable strict mode
1110 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1111 or other complicated routing, then loose mode is recommended.
1113 The max value from conf/{all,interface}/rp_filter is used
1114 when doing source validation on the {interface}.
1116 Default value is 0. Note that some distributions enable it
1119 arp_filter - BOOLEAN
1120 1 - Allows you to have multiple network interfaces on the same
1121 subnet, and have the ARPs for each interface be answered
1122 based on whether or not the kernel would route a packet from
1123 the ARP'd IP out that interface (therefore you must use source
1124 based routing for this to work). In other words it allows control
1125 of which cards (usually 1) will respond to an arp request.
1127 0 - (default) The kernel can respond to arp requests with addresses
1128 from other interfaces. This may seem wrong but it usually makes
1129 sense, because it increases the chance of successful communication.
1130 IP addresses are owned by the complete host on Linux, not by
1131 particular interfaces. Only for more complex setups like load-
1132 balancing, does this behaviour cause problems.
1134 arp_filter for the interface will be enabled if at least one of
1135 conf/{all,interface}/arp_filter is set to TRUE,
1136 it will be disabled otherwise
1138 arp_announce - INTEGER
1139 Define different restriction levels for announcing the local
1140 source IP address from IP packets in ARP requests sent on
1142 0 - (default) Use any local address, configured on any interface
1143 1 - Try to avoid local addresses that are not in the target's
1144 subnet for this interface. This mode is useful when target
1145 hosts reachable via this interface require the source IP
1146 address in ARP requests to be part of their logical network
1147 configured on the receiving interface. When we generate the
1148 request we will check all our subnets that include the
1149 target IP and will preserve the source address if it is from
1150 such subnet. If there is no such subnet we select source
1151 address according to the rules for level 2.
1152 2 - Always use the best local address for this target.
1153 In this mode we ignore the source address in the IP packet
1154 and try to select local address that we prefer for talks with
1155 the target host. Such local address is selected by looking
1156 for primary IP addresses on all our subnets on the outgoing
1157 interface that include the target IP address. If no suitable
1158 local address is found we select the first local address
1159 we have on the outgoing interface or on all other interfaces,
1160 with the hope we will receive reply for our request and
1161 even sometimes no matter the source IP address we announce.
1163 The max value from conf/{all,interface}/arp_announce is used.
1165 Increasing the restriction level gives more chance for
1166 receiving answer from the resolved target while decreasing
1167 the level announces more valid sender's information.
1169 arp_ignore - INTEGER
1170 Define different modes for sending replies in response to
1171 received ARP requests that resolve local target IP addresses:
1172 0 - (default): reply for any local target IP address, configured
1174 1 - reply only if the target IP address is local address
1175 configured on the incoming interface
1176 2 - reply only if the target IP address is local address
1177 configured on the incoming interface and both with the
1178 sender's IP address are part from same subnet on this interface
1179 3 - do not reply for local addresses configured with scope host,
1180 only resolutions for global and link addresses are replied
1182 8 - do not reply for all local addresses
1184 The max value from conf/{all,interface}/arp_ignore is used
1185 when ARP request is received on the {interface}
1187 arp_notify - BOOLEAN
1188 Define mode for notification of address and device changes.
1189 0 - (default): do nothing
1190 1 - Generate gratuitous arp requests when device is brought up
1191 or hardware address changes.
1193 arp_accept - BOOLEAN
1194 Define behavior for gratuitous ARP frames who's IP is not
1195 already present in the ARP table:
1196 0 - don't create new entries in the ARP table
1197 1 - create new entries in the ARP table
1199 Both replies and requests type gratuitous arp will trigger the
1200 ARP table to be updated, if this setting is on.
1202 If the ARP table already contains the IP address of the
1203 gratuitous arp frame, the arp table will be updated regardless
1204 if this setting is on or off.
1206 mcast_solicit - INTEGER
1207 The maximum number of multicast probes in INCOMPLETE state,
1208 when the associated hardware address is unknown. Defaults
1211 ucast_solicit - INTEGER
1212 The maximum number of unicast probes in PROBE state, when
1213 the hardware address is being reconfirmed. Defaults to 3.
1215 app_solicit - INTEGER
1216 The maximum number of probes to send to the user space ARP daemon
1217 via netlink before dropping back to multicast probes (see
1218 mcast_resolicit). Defaults to 0.
1220 mcast_resolicit - INTEGER
1221 The maximum number of multicast probes after unicast and
1222 app probes in PROBE state. Defaults to 0.
1224 disable_policy - BOOLEAN
1225 Disable IPSEC policy (SPD) for this interface
1227 disable_xfrm - BOOLEAN
1228 Disable IPSEC encryption on this interface, whatever the policy
1230 igmpv2_unsolicited_report_interval - INTEGER
1231 The interval in milliseconds in which the next unsolicited
1232 IGMPv1 or IGMPv2 report retransmit will take place.
1233 Default: 10000 (10 seconds)
1235 igmpv3_unsolicited_report_interval - INTEGER
1236 The interval in milliseconds in which the next unsolicited
1237 IGMPv3 report retransmit will take place.
1238 Default: 1000 (1 seconds)
1240 promote_secondaries - BOOLEAN
1241 When a primary IP address is removed from this interface
1242 promote a corresponding secondary IP address instead of
1243 removing all the corresponding secondary IP addresses.
1245 drop_unicast_in_l2_multicast - BOOLEAN
1246 Drop any unicast IP packets that are received in link-layer
1247 multicast (or broadcast) frames.
1248 This behavior (for multicast) is actually a SHOULD in RFC
1249 1122, but is disabled by default for compatibility reasons.
1252 drop_gratuitous_arp - BOOLEAN
1253 Drop all gratuitous ARP frames, for example if there's a known
1254 good ARP proxy on the network and such frames need not be used
1255 (or in the case of 802.11, must not be used to prevent attacks.)
1260 Allows you to write a number, which can be used as required.
1263 xfrm4_gc_thresh - INTEGER
1264 The threshold at which we will start garbage collecting for IPv4
1265 destination cache entries. At twice this value the system will
1266 refuse new allocations. The value must be set below the flowcache
1267 limit (4096 * number of online cpus) to take effect.
1269 igmp_link_local_mcast_reports - BOOLEAN
1270 Enable IGMP reports for link local multicast groups in the
1275 kuznet@ms2.inr.ac.ru
1281 delon.nicolas@wanadoo.fr
1286 /proc/sys/net/ipv6/* Variables:
1288 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1289 apply to IPv6 [XXX?].
1291 bindv6only - BOOLEAN
1292 Default value for IPV6_V6ONLY socket option,
1293 which restricts use of the IPv6 socket to IPv6 communication
1295 TRUE: disable IPv4-mapped address feature
1296 FALSE: enable IPv4-mapped address feature
1298 Default: FALSE (as specified in RFC3493)
1300 flowlabel_consistency - BOOLEAN
1301 Protect the consistency (and unicity) of flow label.
1302 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1308 auto_flowlabels - INTEGER
1309 Automatically generate flow labels based on a flow hash of the
1310 packet. This allows intermediate devices, such as routers, to
1311 identify packet flows for mechanisms like Equal Cost Multipath
1312 Routing (see RFC 6438).
1313 0: automatic flow labels are completely disabled
1314 1: automatic flow labels are enabled by default, they can be
1315 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1317 2: automatic flow labels are allowed, they may be enabled on a
1318 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1319 3: automatic flow labels are enabled and enforced, they cannot
1320 be disabled by the socket option
1323 flowlabel_state_ranges - BOOLEAN
1324 Split the flow label number space into two ranges. 0-0x7FFFF is
1325 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1326 is reserved for stateless flow labels as described in RFC6437.
1331 anycast_src_echo_reply - BOOLEAN
1332 Controls the use of anycast addresses as source addresses for ICMPv6
1338 idgen_delay - INTEGER
1339 Controls the delay in seconds after which time to retry
1340 privacy stable address generation if a DAD conflict is
1342 Default: 1 (as specified in RFC7217)
1344 idgen_retries - INTEGER
1345 Controls the number of retries to generate a stable privacy
1346 address if a DAD conflict is detected.
1347 Default: 3 (as specified in RFC7217)
1350 Controls the MLD query robustness variable (see RFC3810 9.1).
1351 Default: 2 (as specified by RFC3810 9.1)
1352 Minimum: 1 (as specified by RFC6636 4.5)
1356 ip6frag_high_thresh - INTEGER
1357 Maximum memory used to reassemble IPv6 fragments. When
1358 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1359 the fragment handler will toss packets until ip6frag_low_thresh
1362 ip6frag_low_thresh - INTEGER
1363 See ip6frag_high_thresh
1365 ip6frag_time - INTEGER
1366 Time in seconds to keep an IPv6 fragment in memory.
1369 Change the interface-specific default settings.
1373 Change all the interface-specific settings.
1375 [XXX: Other special features than forwarding?]
1377 conf/all/forwarding - BOOLEAN
1378 Enable global IPv6 forwarding between all interfaces.
1380 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1381 to control which interfaces may forward packets and which not.
1383 This also sets all interfaces' Host/Router setting
1384 'forwarding' to the specified value. See below for details.
1386 This referred to as global forwarding.
1391 fwmark_reflect - BOOLEAN
1392 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1393 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1394 If unset, these packets have a fwmark of zero. If set, they have the
1395 fwmark of the packet they are replying to.
1399 Change special settings per interface.
1401 The functional behaviour for certain settings is different
1402 depending on whether local forwarding is enabled or not.
1405 Accept Router Advertisements; autoconfigure using them.
1407 It also determines whether or not to transmit Router
1408 Solicitations. If and only if the functional setting is to
1409 accept Router Advertisements, Router Solicitations will be
1412 Possible values are:
1413 0 Do not accept Router Advertisements.
1414 1 Accept Router Advertisements if forwarding is disabled.
1415 2 Overrule forwarding behaviour. Accept Router Advertisements
1416 even if forwarding is enabled.
1418 Functional default: enabled if local forwarding is disabled.
1419 disabled if local forwarding is enabled.
1421 accept_ra_defrtr - BOOLEAN
1422 Learn default router in Router Advertisement.
1424 Functional default: enabled if accept_ra is enabled.
1425 disabled if accept_ra is disabled.
1427 accept_ra_from_local - BOOLEAN
1428 Accept RA with source-address that is found on local machine
1429 if the RA is otherwise proper and able to be accepted.
1430 Default is to NOT accept these as it may be an un-intended
1434 enabled if accept_ra_from_local is enabled
1435 on a specific interface.
1436 disabled if accept_ra_from_local is disabled
1437 on a specific interface.
1439 accept_ra_min_hop_limit - INTEGER
1440 Minimum hop limit Information in Router Advertisement.
1442 Hop limit Information in Router Advertisement less than this
1443 variable shall be ignored.
1447 accept_ra_pinfo - BOOLEAN
1448 Learn Prefix Information in Router Advertisement.
1450 Functional default: enabled if accept_ra is enabled.
1451 disabled if accept_ra is disabled.
1453 accept_ra_rt_info_max_plen - INTEGER
1454 Maximum prefix length of Route Information in RA.
1456 Route Information w/ prefix larger than or equal to this
1457 variable shall be ignored.
1459 Functional default: 0 if accept_ra_rtr_pref is enabled.
1460 -1 if accept_ra_rtr_pref is disabled.
1462 accept_ra_rtr_pref - BOOLEAN
1463 Accept Router Preference in RA.
1465 Functional default: enabled if accept_ra is enabled.
1466 disabled if accept_ra is disabled.
1468 accept_ra_mtu - BOOLEAN
1469 Apply the MTU value specified in RA option 5 (RFC4861). If
1470 disabled, the MTU specified in the RA will be ignored.
1472 Functional default: enabled if accept_ra is enabled.
1473 disabled if accept_ra is disabled.
1475 accept_redirects - BOOLEAN
1478 Functional default: enabled if local forwarding is disabled.
1479 disabled if local forwarding is enabled.
1481 accept_source_route - INTEGER
1482 Accept source routing (routing extension header).
1484 >= 0: Accept only routing header type 2.
1485 < 0: Do not accept routing header.
1490 Autoconfigure addresses using Prefix Information in Router
1493 Functional default: enabled if accept_ra_pinfo is enabled.
1494 disabled if accept_ra_pinfo is disabled.
1496 dad_transmits - INTEGER
1497 The amount of Duplicate Address Detection probes to send.
1500 forwarding - INTEGER
1501 Configure interface-specific Host/Router behaviour.
1503 Note: It is recommended to have the same setting on all
1504 interfaces; mixed router/host scenarios are rather uncommon.
1506 Possible values are:
1507 0 Forwarding disabled
1508 1 Forwarding enabled
1512 By default, Host behaviour is assumed. This means:
1514 1. IsRouter flag is not set in Neighbour Advertisements.
1515 2. If accept_ra is TRUE (default), transmit Router
1517 3. If accept_ra is TRUE (default), accept Router
1518 Advertisements (and do autoconfiguration).
1519 4. If accept_redirects is TRUE (default), accept Redirects.
1523 If local forwarding is enabled, Router behaviour is assumed.
1524 This means exactly the reverse from the above:
1526 1. IsRouter flag is set in Neighbour Advertisements.
1527 2. Router Solicitations are not sent unless accept_ra is 2.
1528 3. Router Advertisements are ignored unless accept_ra is 2.
1529 4. Redirects are ignored.
1531 Default: 0 (disabled) if global forwarding is disabled (default),
1532 otherwise 1 (enabled).
1535 Default Hop Limit to set.
1539 Default Maximum Transfer Unit
1540 Default: 1280 (IPv6 required minimum)
1542 ip_nonlocal_bind - BOOLEAN
1543 If set, allows processes to bind() to non-local IPv6 addresses,
1544 which can be quite useful - but may break some applications.
1547 router_probe_interval - INTEGER
1548 Minimum interval (in seconds) between Router Probing described
1553 router_solicitation_delay - INTEGER
1554 Number of seconds to wait after interface is brought up
1555 before sending Router Solicitations.
1558 router_solicitation_interval - INTEGER
1559 Number of seconds to wait between Router Solicitations.
1562 router_solicitations - INTEGER
1563 Number of Router Solicitations to send until assuming no
1564 routers are present.
1567 use_oif_addrs_only - BOOLEAN
1568 When enabled, the candidate source addresses for destinations
1569 routed via this interface are restricted to the set of addresses
1570 configured on this interface (vis. RFC 6724, section 4).
1574 use_tempaddr - INTEGER
1575 Preference for Privacy Extensions (RFC3041).
1576 <= 0 : disable Privacy Extensions
1577 == 1 : enable Privacy Extensions, but prefer public
1578 addresses over temporary addresses.
1579 > 1 : enable Privacy Extensions and prefer temporary
1580 addresses over public addresses.
1581 Default: 0 (for most devices)
1582 -1 (for point-to-point devices and loopback devices)
1584 temp_valid_lft - INTEGER
1585 valid lifetime (in seconds) for temporary addresses.
1586 Default: 604800 (7 days)
1588 temp_prefered_lft - INTEGER
1589 Preferred lifetime (in seconds) for temporary addresses.
1590 Default: 86400 (1 day)
1592 keep_addr_on_down - INTEGER
1593 Keep all IPv6 addresses on an interface down event. If set static
1594 global addresses with no expiration time are not flushed.
1599 Default: 0 (addresses are removed)
1601 max_desync_factor - INTEGER
1602 Maximum value for DESYNC_FACTOR, which is a random value
1603 that ensures that clients don't synchronize with each
1604 other and generate new addresses at exactly the same time.
1605 value is in seconds.
1608 regen_max_retry - INTEGER
1609 Number of attempts before give up attempting to generate
1610 valid temporary addresses.
1613 max_addresses - INTEGER
1614 Maximum number of autoconfigured addresses per interface. Setting
1615 to zero disables the limitation. It is not recommended to set this
1616 value too large (or to zero) because it would be an easy way to
1617 crash the kernel by allowing too many addresses to be created.
1620 disable_ipv6 - BOOLEAN
1621 Disable IPv6 operation. If accept_dad is set to 2, this value
1622 will be dynamically set to TRUE if DAD fails for the link-local
1624 Default: FALSE (enable IPv6 operation)
1626 When this value is changed from 1 to 0 (IPv6 is being enabled),
1627 it will dynamically create a link-local address on the given
1628 interface and start Duplicate Address Detection, if necessary.
1630 When this value is changed from 0 to 1 (IPv6 is being disabled),
1631 it will dynamically delete all address on the given interface.
1633 accept_dad - INTEGER
1634 Whether to accept DAD (Duplicate Address Detection).
1636 1: Enable DAD (default)
1637 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1638 link-local address has been found.
1640 force_tllao - BOOLEAN
1641 Enable sending the target link-layer address option even when
1642 responding to a unicast neighbor solicitation.
1645 Quoting from RFC 2461, section 4.4, Target link-layer address:
1647 "The option MUST be included for multicast solicitations in order to
1648 avoid infinite Neighbor Solicitation "recursion" when the peer node
1649 does not have a cache entry to return a Neighbor Advertisements
1650 message. When responding to unicast solicitations, the option can be
1651 omitted since the sender of the solicitation has the correct link-
1652 layer address; otherwise it would not have be able to send the unicast
1653 solicitation in the first place. However, including the link-layer
1654 address in this case adds little overhead and eliminates a potential
1655 race condition where the sender deletes the cached link-layer address
1656 prior to receiving a response to a previous solicitation."
1658 ndisc_notify - BOOLEAN
1659 Define mode for notification of address and device changes.
1660 0 - (default): do nothing
1661 1 - Generate unsolicited neighbour advertisements when device is brought
1662 up or hardware address changes.
1664 mldv1_unsolicited_report_interval - INTEGER
1665 The interval in milliseconds in which the next unsolicited
1666 MLDv1 report retransmit will take place.
1667 Default: 10000 (10 seconds)
1669 mldv2_unsolicited_report_interval - INTEGER
1670 The interval in milliseconds in which the next unsolicited
1671 MLDv2 report retransmit will take place.
1672 Default: 1000 (1 second)
1674 force_mld_version - INTEGER
1675 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1676 1 - Enforce to use MLD version 1
1677 2 - Enforce to use MLD version 2
1679 suppress_frag_ndisc - INTEGER
1680 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1681 with IPv6 Neighbor Discovery) behavior:
1682 1 - (default) discard fragmented neighbor discovery packets
1683 0 - allow fragmented neighbor discovery packets
1685 optimistic_dad - BOOLEAN
1686 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1687 0: disabled (default)
1690 use_optimistic - BOOLEAN
1691 If enabled, do not classify optimistic addresses as deprecated during
1692 source address selection. Preferred addresses will still be chosen
1693 before optimistic addresses, subject to other ranking in the source
1694 address selection algorithm.
1695 0: disabled (default)
1698 stable_secret - IPv6 address
1699 This IPv6 address will be used as a secret to generate IPv6
1700 addresses for link-local addresses and autoconfigured
1701 ones. All addresses generated after setting this secret will
1702 be stable privacy ones by default. This can be changed via the
1703 addrgenmode ip-link. conf/default/stable_secret is used as the
1704 secret for the namespace, the interface specific ones can
1705 overwrite that. Writes to conf/all/stable_secret are refused.
1707 It is recommended to generate this secret during installation
1708 of a system and keep it stable after that.
1710 By default the stable secret is unset.
1712 drop_unicast_in_l2_multicast - BOOLEAN
1713 Drop any unicast IPv6 packets that are received in link-layer
1714 multicast (or broadcast) frames.
1716 By default this is turned off.
1718 drop_unsolicited_na - BOOLEAN
1719 Drop all unsolicited neighbor advertisements, for example if there's
1720 a known good NA proxy on the network and such frames need not be used
1721 (or in the case of 802.11, must not be used to prevent attacks.)
1723 By default this is turned off.
1725 enhanced_dad - BOOLEAN
1726 Include a nonce option in the IPv6 neighbor solicitation messages used for
1727 duplicate address detection per RFC7527. A received DAD NS will only signal
1728 a duplicate address if the nonce is different. This avoids any false
1729 detection of duplicates due to loopback of the NS messages that we send.
1730 The nonce option will be sent on an interface unless both of
1731 conf/{all,interface}/enhanced_dad are set to FALSE.
1736 Limit the maximal rates for sending ICMPv6 packets.
1737 0 to disable any limiting,
1738 otherwise the minimal space between responses in milliseconds.
1741 xfrm6_gc_thresh - INTEGER
1742 The threshold at which we will start garbage collecting for IPv6
1743 destination cache entries. At twice this value the system will
1744 refuse new allocations. The value must be set below the flowcache
1745 limit (4096 * number of online cpus) to take effect.
1749 Pekka Savola <pekkas@netcore.fi>
1750 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1753 /proc/sys/net/bridge/* Variables:
1755 bridge-nf-call-arptables - BOOLEAN
1756 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1760 bridge-nf-call-iptables - BOOLEAN
1761 1 : pass bridged IPv4 traffic to iptables' chains.
1765 bridge-nf-call-ip6tables - BOOLEAN
1766 1 : pass bridged IPv6 traffic to ip6tables' chains.
1770 bridge-nf-filter-vlan-tagged - BOOLEAN
1771 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1775 bridge-nf-filter-pppoe-tagged - BOOLEAN
1776 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1780 bridge-nf-pass-vlan-input-dev - BOOLEAN
1781 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1782 interface on the bridge and set the netfilter input device to the vlan.
1783 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1784 target work with vlan-on-top-of-bridge interfaces. When no matching
1785 vlan interface is found, or this switch is off, the input device is
1786 set to the bridge interface.
1787 0: disable bridge netfilter vlan interface lookup.
1790 proc/sys/net/sctp/* Variables:
1792 addip_enable - BOOLEAN
1793 Enable or disable extension of Dynamic Address Reconfiguration
1794 (ADD-IP) functionality specified in RFC5061. This extension provides
1795 the ability to dynamically add and remove new addresses for the SCTP
1798 1: Enable extension.
1800 0: Disable extension.
1805 Enable or disable pf (pf is short for potentially failed) state. A value
1806 of pf_retrans > path_max_retrans also disables pf state. That is, one of
1807 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
1808 Since pf_retrans and path_max_retrans can be changed by userspace
1809 application, sometimes user expects to disable pf state by the value of
1810 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
1811 or path_max_retrans is changed by the user application, this pf state is
1812 enabled. As such, it is necessary to add this to dynamically enable
1813 and disable pf state. See:
1814 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
1823 addip_noauth_enable - BOOLEAN
1824 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1825 authentication to protect the operations of adding or removing new
1826 addresses. This requirement is mandated so that unauthorized hosts
1827 would not be able to hijack associations. However, older
1828 implementations may not have implemented this requirement while
1829 allowing the ADD-IP extension. For reasons of interoperability,
1830 we provide this variable to control the enforcement of the
1831 authentication requirement.
1833 1: Allow ADD-IP extension to be used without authentication. This
1834 should only be set in a closed environment for interoperability
1835 with older implementations.
1837 0: Enforce the authentication requirement
1841 auth_enable - BOOLEAN
1842 Enable or disable Authenticated Chunks extension. This extension
1843 provides the ability to send and receive authenticated chunks and is
1844 required for secure operation of Dynamic Address Reconfiguration
1847 1: Enable this extension.
1848 0: Disable this extension.
1852 prsctp_enable - BOOLEAN
1853 Enable or disable the Partial Reliability extension (RFC3758) which
1854 is used to notify peers that a given DATA should no longer be expected.
1862 The limit of the number of new packets that can be initially sent. It
1863 controls how bursty the generated traffic can be.
1867 association_max_retrans - INTEGER
1868 Set the maximum number for retransmissions that an association can
1869 attempt deciding that the remote end is unreachable. If this value
1870 is exceeded, the association is terminated.
1874 max_init_retransmits - INTEGER
1875 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1876 that an association will attempt before declaring the destination
1877 unreachable and terminating.
1881 path_max_retrans - INTEGER
1882 The maximum number of retransmissions that will be attempted on a given
1883 path. Once this threshold is exceeded, the path is considered
1884 unreachable, and new traffic will use a different path when the
1885 association is multihomed.
1889 pf_retrans - INTEGER
1890 The number of retransmissions that will be attempted on a given path
1891 before traffic is redirected to an alternate transport (should one
1892 exist). Note this is distinct from path_max_retrans, as a path that
1893 passes the pf_retrans threshold can still be used. Its only
1894 deprioritized when a transmission path is selected by the stack. This
1895 setting is primarily used to enable fast failover mechanisms without
1896 having to reduce path_max_retrans to a very low value. See:
1897 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1898 for details. Note also that a value of pf_retrans > path_max_retrans
1899 disables this feature. Since both pf_retrans and path_max_retrans can
1900 be changed by userspace application, a variable pf_enable is used to
1905 rto_initial - INTEGER
1906 The initial round trip timeout value in milliseconds that will be used
1907 in calculating round trip times. This is the initial time interval
1908 for retransmissions.
1913 The maximum value (in milliseconds) of the round trip timeout. This
1914 is the largest time interval that can elapse between retransmissions.
1919 The minimum value (in milliseconds) of the round trip timeout. This
1920 is the smallest time interval the can elapse between retransmissions.
1924 hb_interval - INTEGER
1925 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1926 are sent at the specified interval on idle paths to probe the state of
1927 a given path between 2 associations.
1931 sack_timeout - INTEGER
1932 The amount of time (in milliseconds) that the implementation will wait
1937 valid_cookie_life - INTEGER
1938 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1939 is used during association establishment.
1943 cookie_preserve_enable - BOOLEAN
1944 Enable or disable the ability to extend the lifetime of the SCTP cookie
1945 that is used during the establishment phase of SCTP association
1947 1: Enable cookie lifetime extension.
1952 cookie_hmac_alg - STRING
1953 Select the hmac algorithm used when generating the cookie value sent by
1954 a listening sctp socket to a connecting client in the INIT-ACK chunk.
1959 Ability to assign md5 or sha1 as the selected alg is predicated on the
1960 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
1961 CONFIG_CRYPTO_SHA1).
1963 Default: Dependent on configuration. MD5 if available, else SHA1 if
1964 available, else none.
1966 rcvbuf_policy - INTEGER
1967 Determines if the receive buffer is attributed to the socket or to
1968 association. SCTP supports the capability to create multiple
1969 associations on a single socket. When using this capability, it is
1970 possible that a single stalled association that's buffering a lot
1971 of data may block other associations from delivering their data by
1972 consuming all of the receive buffer space. To work around this,
1973 the rcvbuf_policy could be set to attribute the receiver buffer space
1974 to each association instead of the socket. This prevents the described
1977 1: rcvbuf space is per association
1978 0: rcvbuf space is per socket
1982 sndbuf_policy - INTEGER
1983 Similar to rcvbuf_policy above, this applies to send buffer space.
1985 1: Send buffer is tracked per association
1986 0: Send buffer is tracked per socket.
1990 sctp_mem - vector of 3 INTEGERs: min, pressure, max
1991 Number of pages allowed for queueing by all SCTP sockets.
1993 min: Below this number of pages SCTP is not bothered about its
1994 memory appetite. When amount of memory allocated by SCTP exceeds
1995 this number, SCTP starts to moderate memory usage.
1997 pressure: This value was introduced to follow format of tcp_mem.
1999 max: Number of pages allowed for queueing by all SCTP sockets.
2001 Default is calculated at boot time from amount of available memory.
2003 sctp_rmem - vector of 3 INTEGERs: min, default, max
2004 Only the first value ("min") is used, "default" and "max" are
2007 min: Minimal size of receive buffer used by SCTP socket.
2008 It is guaranteed to each SCTP socket (but not association) even
2009 under moderate memory pressure.
2013 sctp_wmem - vector of 3 INTEGERs: min, default, max
2014 Currently this tunable has no effect.
2016 addr_scope_policy - INTEGER
2017 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2019 0 - Disable IPv4 address scoping
2020 1 - Enable IPv4 address scoping
2021 2 - Follow draft but allow IPv4 private addresses
2022 3 - Follow draft but allow IPv4 link local addresses
2027 /proc/sys/net/core/*
2028 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
2031 /proc/sys/net/unix/*
2032 max_dgram_qlen - INTEGER
2033 The maximum length of dgram socket receive queue
2040 /proc/sys/net/irda/*
2041 fast_poll_increase FIXME
2042 warn_noreply_time FIXME
2043 discovery_slots FIXME
2046 discovery_timeout FIXME
2047 lap_keepalive_time FIXME
2048 max_noreply_time FIXME
2049 max_tx_data_size FIXME
2051 min_tx_turn_time FIXME