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 - BOOLEAN
19 Disable Path MTU Discovery.
23 default 552 - minimum discovered Path MTU
25 route/max_size - INTEGER
26 Maximum number of routes allowed in the kernel. Increase
27 this when using large numbers of interfaces and/or routes.
29 neigh/default/gc_thresh1 - INTEGER
30 Minimum number of entries to keep. Garbage collector will not
31 purge entries if there are fewer than this number.
34 neigh/default/gc_thresh3 - INTEGER
35 Maximum number of neighbor entries allowed. Increase this
36 when using large numbers of interfaces and when communicating
37 with large numbers of directly-connected peers.
40 neigh/default/unres_qlen_bytes - INTEGER
41 The maximum number of bytes which may be used by packets
42 queued for each unresolved address by other network layers.
44 Setting negative value is meaningless and will return error.
45 Default: 65536 Bytes(64KB)
47 neigh/default/unres_qlen - INTEGER
48 The maximum number of packets which may be queued for each
49 unresolved address by other network layers.
50 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
51 Prior to linux 3.3, the default value is 3 which may cause
52 unexpected packet loss. The current default value is calculated
53 according to default value of unres_qlen_bytes and true size of
58 Time, in seconds, that cached PMTU information is kept.
61 The advertised MSS depends on the first hop route MTU, but will
62 never be lower than this setting.
66 ipfrag_high_thresh - INTEGER
67 Maximum memory used to reassemble IP fragments. When
68 ipfrag_high_thresh bytes of memory is allocated for this purpose,
69 the fragment handler will toss packets until ipfrag_low_thresh
72 ipfrag_low_thresh - INTEGER
73 See ipfrag_high_thresh
76 Time in seconds to keep an IP fragment in memory.
78 ipfrag_secret_interval - INTEGER
79 Regeneration interval (in seconds) of the hash secret (or lifetime
80 for the hash secret) for IP fragments.
83 ipfrag_max_dist - INTEGER
84 ipfrag_max_dist is a non-negative integer value which defines the
85 maximum "disorder" which is allowed among fragments which share a
86 common IP source address. Note that reordering of packets is
87 not unusual, but if a large number of fragments arrive from a source
88 IP address while a particular fragment queue remains incomplete, it
89 probably indicates that one or more fragments belonging to that queue
90 have been lost. When ipfrag_max_dist is positive, an additional check
91 is done on fragments before they are added to a reassembly queue - if
92 ipfrag_max_dist (or more) fragments have arrived from a particular IP
93 address between additions to any IP fragment queue using that source
94 address, it's presumed that one or more fragments in the queue are
95 lost. The existing fragment queue will be dropped, and a new one
96 started. An ipfrag_max_dist value of zero disables this check.
98 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
99 result in unnecessarily dropping fragment queues when normal
100 reordering of packets occurs, which could lead to poor application
101 performance. Using a very large value, e.g. 50000, increases the
102 likelihood of incorrectly reassembling IP fragments that originate
103 from different IP datagrams, which could result in data corruption.
108 inet_peer_threshold - INTEGER
109 The approximate size of the storage. Starting from this threshold
110 entries will be thrown aggressively. This threshold also determines
111 entries' time-to-live and time intervals between garbage collection
112 passes. More entries, less time-to-live, less GC interval.
114 inet_peer_minttl - INTEGER
115 Minimum time-to-live of entries. Should be enough to cover fragment
116 time-to-live on the reassembling side. This minimum time-to-live is
117 guaranteed if the pool size is less than inet_peer_threshold.
120 inet_peer_maxttl - INTEGER
121 Maximum time-to-live of entries. Unused entries will expire after
122 this period of time if there is no memory pressure on the pool (i.e.
123 when the number of entries in the pool is very small).
129 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
130 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
133 tcp_abort_on_overflow - BOOLEAN
134 If listening service is too slow to accept new connections,
135 reset them. Default state is FALSE. It means that if overflow
136 occurred due to a burst, connection will recover. Enable this
137 option _only_ if you are really sure that listening daemon
138 cannot be tuned to accept connections faster. Enabling this
139 option can harm clients of your server.
141 tcp_adv_win_scale - INTEGER
142 Count buffering overhead as bytes/2^tcp_adv_win_scale
143 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
145 Possible values are [-31, 31], inclusive.
148 tcp_allowed_congestion_control - STRING
149 Show/set the congestion control choices available to non-privileged
150 processes. The list is a subset of those listed in
151 tcp_available_congestion_control.
152 Default is "reno" and the default setting (tcp_congestion_control).
154 tcp_app_win - INTEGER
155 Reserve max(window/2^tcp_app_win, mss) of window for application
156 buffer. Value 0 is special, it means that nothing is reserved.
159 tcp_available_congestion_control - STRING
160 Shows the available congestion control choices that are registered.
161 More congestion control algorithms may be available as modules,
164 tcp_base_mss - INTEGER
165 The initial value of search_low to be used by the packetization layer
166 Path MTU discovery (MTU probing). If MTU probing is enabled,
167 this is the initial MSS used by the connection.
169 tcp_congestion_control - STRING
170 Set the congestion control algorithm to be used for new
171 connections. The algorithm "reno" is always available, but
172 additional choices may be available based on kernel configuration.
173 Default is set as part of kernel configuration.
174 For passive connections, the listener congestion control choice
176 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
179 Allows TCP to send "duplicate" SACKs.
181 tcp_early_retrans - INTEGER
182 Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
183 for triggering fast retransmit when the amount of outstanding data is
184 small and when no previously unsent data can be transmitted (such
185 that limited transmit could be used). Also controls the use of
186 Tail loss probe (TLP) that converts RTOs occurring due to tail
187 losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
191 2 enables ER but delays fast recovery and fast retransmit
192 by a fourth of RTT. This mitigates connection falsely
193 recovers when network has a small degree of reordering
194 (less than 3 packets).
195 3 enables delayed ER and TLP.
200 Control use of Explicit Congestion Notification (ECN) by TCP.
201 ECN is used only when both ends of the TCP connection indicate
202 support for it. This feature is useful in avoiding losses due
203 to congestion by allowing supporting routers to signal
204 congestion before having to drop packets.
206 0 Disable ECN. Neither initiate nor accept ECN.
207 1 Enable ECN when requested by incoming connections and
208 also request ECN on outgoing connection attempts.
209 2 Enable ECN when requested by incoming connections
210 but do not request ECN on outgoing connections.
214 Enable FACK congestion avoidance and fast retransmission.
215 The value is not used, if tcp_sack is not enabled.
217 tcp_fin_timeout - INTEGER
218 The length of time an orphaned (no longer referenced by any
219 application) connection will remain in the FIN_WAIT_2 state
220 before it is aborted at the local end. While a perfectly
221 valid "receive only" state for an un-orphaned connection, an
222 orphaned connection in FIN_WAIT_2 state could otherwise wait
223 forever for the remote to close its end of the connection.
228 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
229 F-RTO is an enhanced recovery algorithm for TCP retransmission
230 timeouts. It is particularly beneficial in networks where the
231 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
232 modification. It does not require any support from the peer.
234 By default it's enabled with a non-zero value. 0 disables F-RTO.
236 tcp_keepalive_time - INTEGER
237 How often TCP sends out keepalive messages when keepalive is enabled.
240 tcp_keepalive_probes - INTEGER
241 How many keepalive probes TCP sends out, until it decides that the
242 connection is broken. Default value: 9.
244 tcp_keepalive_intvl - INTEGER
245 How frequently the probes are send out. Multiplied by
246 tcp_keepalive_probes it is time to kill not responding connection,
247 after probes started. Default value: 75sec i.e. connection
248 will be aborted after ~11 minutes of retries.
250 tcp_low_latency - BOOLEAN
251 If set, the TCP stack makes decisions that prefer lower
252 latency as opposed to higher throughput. By default, this
253 option is not set meaning that higher throughput is preferred.
254 An example of an application where this default should be
255 changed would be a Beowulf compute cluster.
258 tcp_max_orphans - INTEGER
259 Maximal number of TCP sockets not attached to any user file handle,
260 held by system. If this number is exceeded orphaned connections are
261 reset immediately and warning is printed. This limit exists
262 only to prevent simple DoS attacks, you _must_ not rely on this
263 or lower the limit artificially, but rather increase it
264 (probably, after increasing installed memory),
265 if network conditions require more than default value,
266 and tune network services to linger and kill such states
267 more aggressively. Let me to remind again: each orphan eats
268 up to ~64K of unswappable memory.
270 tcp_max_ssthresh - INTEGER
271 Limited Slow-Start for TCP with large congestion windows (cwnd) defined in
272 RFC3742. Limited slow-start is a mechanism to limit growth of the cwnd
273 on the region where cwnd is larger than tcp_max_ssthresh. TCP increases cwnd
274 by at most tcp_max_ssthresh segments, and by at least tcp_max_ssthresh/2
275 segments per RTT when the cwnd is above tcp_max_ssthresh.
276 If TCP connection increased cwnd to thousands (or tens of thousands) segments,
277 and thousands of packets were being dropped during slow-start, you can set
278 tcp_max_ssthresh to improve performance for new TCP connection.
281 tcp_max_syn_backlog - INTEGER
282 Maximal number of remembered connection requests, which have not
283 received an acknowledgment from connecting client.
284 The minimal value is 128 for low memory machines, and it will
285 increase in proportion to the memory of machine.
286 If server suffers from overload, try increasing this number.
288 tcp_max_tw_buckets - INTEGER
289 Maximal number of timewait sockets held by system simultaneously.
290 If this number is exceeded time-wait socket is immediately destroyed
291 and warning is printed. This limit exists only to prevent
292 simple DoS attacks, you _must_ not lower the limit artificially,
293 but rather increase it (probably, after increasing installed memory),
294 if network conditions require more than default value.
296 tcp_mem - vector of 3 INTEGERs: min, pressure, max
297 min: below this number of pages TCP is not bothered about its
300 pressure: when amount of memory allocated by TCP exceeds this number
301 of pages, TCP moderates its memory consumption and enters memory
302 pressure mode, which is exited when memory consumption falls
305 max: number of pages allowed for queueing by all TCP sockets.
307 Defaults are calculated at boot time from amount of available
310 tcp_moderate_rcvbuf - BOOLEAN
311 If set, TCP performs receive buffer auto-tuning, attempting to
312 automatically size the buffer (no greater than tcp_rmem[2]) to
313 match the size required by the path for full throughput. Enabled by
316 tcp_mtu_probing - INTEGER
317 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
320 1 - Disabled by default, enabled when an ICMP black hole detected
321 2 - Always enabled, use initial MSS of tcp_base_mss.
323 tcp_no_metrics_save - BOOLEAN
324 By default, TCP saves various connection metrics in the route cache
325 when the connection closes, so that connections established in the
326 near future can use these to set initial conditions. Usually, this
327 increases overall performance, but may sometimes cause performance
328 degradation. If set, TCP will not cache metrics on closing
331 tcp_orphan_retries - INTEGER
332 This value influences the timeout of a locally closed TCP connection,
333 when RTO retransmissions remain unacknowledged.
334 See tcp_retries2 for more details.
336 The default value is 8.
337 If your machine is a loaded WEB server,
338 you should think about lowering this value, such sockets
339 may consume significant resources. Cf. tcp_max_orphans.
341 tcp_reordering - INTEGER
342 Maximal reordering of packets in a TCP stream.
345 tcp_retrans_collapse - BOOLEAN
346 Bug-to-bug compatibility with some broken printers.
347 On retransmit try to send bigger packets to work around bugs in
350 tcp_retries1 - INTEGER
351 This value influences the time, after which TCP decides, that
352 something is wrong due to unacknowledged RTO retransmissions,
353 and reports this suspicion to the network layer.
354 See tcp_retries2 for more details.
356 RFC 1122 recommends at least 3 retransmissions, which is the
359 tcp_retries2 - INTEGER
360 This value influences the timeout of an alive TCP connection,
361 when RTO retransmissions remain unacknowledged.
362 Given a value of N, a hypothetical TCP connection following
363 exponential backoff with an initial RTO of TCP_RTO_MIN would
364 retransmit N times before killing the connection at the (N+1)th RTO.
366 The default value of 15 yields a hypothetical timeout of 924.6
367 seconds and is a lower bound for the effective timeout.
368 TCP will effectively time out at the first RTO which exceeds the
369 hypothetical timeout.
371 RFC 1122 recommends at least 100 seconds for the timeout,
372 which corresponds to a value of at least 8.
374 tcp_rfc1337 - BOOLEAN
375 If set, the TCP stack behaves conforming to RFC1337. If unset,
376 we are not conforming to RFC, but prevent TCP TIME_WAIT
380 tcp_rmem - vector of 3 INTEGERs: min, default, max
381 min: Minimal size of receive buffer used by TCP sockets.
382 It is guaranteed to each TCP socket, even under moderate memory
386 default: initial size of receive buffer used by TCP sockets.
387 This value overrides net.core.rmem_default used by other protocols.
388 Default: 87380 bytes. This value results in window of 65535 with
389 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
390 less for default tcp_app_win. See below about these variables.
392 max: maximal size of receive buffer allowed for automatically
393 selected receiver buffers for TCP socket. This value does not override
394 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
395 automatic tuning of that socket's receive buffer size, in which
396 case this value is ignored.
397 Default: between 87380B and 6MB, depending on RAM size.
400 Enable select acknowledgments (SACKS).
402 tcp_slow_start_after_idle - BOOLEAN
403 If set, provide RFC2861 behavior and time out the congestion
404 window after an idle period. An idle period is defined at
405 the current RTO. If unset, the congestion window will not
406 be timed out after an idle period.
410 Use the Host requirements interpretation of the TCP urgent pointer field.
411 Most hosts use the older BSD interpretation, so if you turn this on
412 Linux might not communicate correctly with them.
415 tcp_synack_retries - INTEGER
416 Number of times SYNACKs for a passive TCP connection attempt will
417 be retransmitted. Should not be higher than 255. Default value
418 is 5, which corresponds to 31seconds till the last retransmission
419 with the current initial RTO of 1second. With this the final timeout
420 for a passive TCP connection will happen after 63seconds.
422 tcp_syncookies - BOOLEAN
423 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
424 Send out syncookies when the syn backlog queue of a socket
425 overflows. This is to prevent against the common 'SYN flood attack'
428 Note, that syncookies is fallback facility.
429 It MUST NOT be used to help highly loaded servers to stand
430 against legal connection rate. If you see SYN flood warnings
431 in your logs, but investigation shows that they occur
432 because of overload with legal connections, you should tune
433 another parameters until this warning disappear.
434 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
436 syncookies seriously violate TCP protocol, do not allow
437 to use TCP extensions, can result in serious degradation
438 of some services (f.e. SMTP relaying), visible not by you,
439 but your clients and relays, contacting you. While you see
440 SYN flood warnings in logs not being really flooded, your server
441 is seriously misconfigured.
443 If you want to test which effects syncookies have to your
444 network connections you can set this knob to 2 to enable
445 unconditionally generation of syncookies.
447 tcp_fastopen - INTEGER
448 Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
449 in the opening SYN packet. To use this feature, the client application
450 must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
451 connect() to perform a TCP handshake automatically.
453 The values (bitmap) are
454 1: Enables sending data in the opening SYN on the client.
455 2: Enables TCP Fast Open on the server side, i.e., allowing data in
456 a SYN packet to be accepted and passed to the application before
457 3-way hand shake finishes.
458 4: Send data in the opening SYN regardless of cookie availability and
459 without a cookie option.
460 0x100: Accept SYN data w/o validating the cookie.
461 0x200: Accept data-in-SYN w/o any cookie option present.
462 0x400/0x800: Enable Fast Open on all listeners regardless of the
463 TCP_FASTOPEN socket option. The two different flags designate two
464 different ways of setting max_qlen without the TCP_FASTOPEN socket
469 Note that the client & server side Fast Open flags (1 and 2
470 respectively) must be also enabled before the rest of flags can take
473 See include/net/tcp.h and the code for more details.
475 tcp_syn_retries - INTEGER
476 Number of times initial SYNs for an active TCP connection attempt
477 will be retransmitted. Should not be higher than 255. Default value
478 is 6, which corresponds to 63seconds till the last retransmission
479 with the current initial RTO of 1second. With this the final timeout
480 for an active TCP connection attempt will happen after 127seconds.
482 tcp_timestamps - BOOLEAN
483 Enable timestamps as defined in RFC1323.
485 tcp_tso_win_divisor - INTEGER
486 This allows control over what percentage of the congestion window
487 can be consumed by a single TSO frame.
488 The setting of this parameter is a choice between burstiness and
489 building larger TSO frames.
492 tcp_tw_recycle - BOOLEAN
493 Enable fast recycling TIME-WAIT sockets. Default value is 0.
494 It should not be changed without advice/request of technical
497 tcp_tw_reuse - BOOLEAN
498 Allow to reuse TIME-WAIT sockets for new connections when it is
499 safe from protocol viewpoint. Default value is 0.
500 It should not be changed without advice/request of technical
503 tcp_window_scaling - BOOLEAN
504 Enable window scaling as defined in RFC1323.
506 tcp_wmem - vector of 3 INTEGERs: min, default, max
507 min: Amount of memory reserved for send buffers for TCP sockets.
508 Each TCP socket has rights to use it due to fact of its birth.
511 default: initial size of send buffer used by TCP sockets. This
512 value overrides net.core.wmem_default used by other protocols.
513 It is usually lower than net.core.wmem_default.
516 max: Maximal amount of memory allowed for automatically tuned
517 send buffers for TCP sockets. This value does not override
518 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
519 automatic tuning of that socket's send buffer size, in which case
520 this value is ignored.
521 Default: between 64K and 4MB, depending on RAM size.
523 tcp_notsent_lowat - UNSIGNED INTEGER
524 A TCP socket can control the amount of unsent bytes in its write queue,
525 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
526 reports POLLOUT events if the amount of unsent bytes is below a per
527 socket value, and if the write queue is not full. sendmsg() will
528 also not add new buffers if the limit is hit.
530 This global variable controls the amount of unsent data for
531 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
532 to the global variable has immediate effect.
534 Default: UINT_MAX (0xFFFFFFFF)
536 tcp_workaround_signed_windows - BOOLEAN
537 If set, assume no receipt of a window scaling option means the
538 remote TCP is broken and treats the window as a signed quantity.
539 If unset, assume the remote TCP is not broken even if we do
540 not receive a window scaling option from them.
543 tcp_dma_copybreak - INTEGER
544 Lower limit, in bytes, of the size of socket reads that will be
545 offloaded to a DMA copy engine, if one is present in the system
546 and CONFIG_NET_DMA is enabled.
549 tcp_thin_linear_timeouts - BOOLEAN
550 Enable dynamic triggering of linear timeouts for thin streams.
551 If set, a check is performed upon retransmission by timeout to
552 determine if the stream is thin (less than 4 packets in flight).
553 As long as the stream is found to be thin, up to 6 linear
554 timeouts may be performed before exponential backoff mode is
555 initiated. This improves retransmission latency for
556 non-aggressive thin streams, often found to be time-dependent.
557 For more information on thin streams, see
558 Documentation/networking/tcp-thin.txt
561 tcp_thin_dupack - BOOLEAN
562 Enable dynamic triggering of retransmissions after one dupACK
563 for thin streams. If set, a check is performed upon reception
564 of a dupACK to determine if the stream is thin (less than 4
565 packets in flight). As long as the stream is found to be thin,
566 data is retransmitted on the first received dupACK. This
567 improves retransmission latency for non-aggressive thin
568 streams, often found to be time-dependent.
569 For more information on thin streams, see
570 Documentation/networking/tcp-thin.txt
573 tcp_limit_output_bytes - INTEGER
574 Controls TCP Small Queue limit per tcp socket.
575 TCP bulk sender tends to increase packets in flight until it
576 gets losses notifications. With SNDBUF autotuning, this can
577 result in a large amount of packets queued in qdisc/device
578 on the local machine, hurting latency of other flows, for
579 typical pfifo_fast qdiscs.
580 tcp_limit_output_bytes limits the number of bytes on qdisc
581 or device to reduce artificial RTT/cwnd and reduce bufferbloat.
582 Note: For GSO/TSO enabled flows, we try to have at least two
583 packets in flight. Reducing tcp_limit_output_bytes might also
584 reduce the size of individual GSO packet (64KB being the max)
587 tcp_challenge_ack_limit - INTEGER
588 Limits number of Challenge ACK sent per second, as recommended
589 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
594 udp_mem - vector of 3 INTEGERs: min, pressure, max
595 Number of pages allowed for queueing by all UDP sockets.
597 min: Below this number of pages UDP is not bothered about its
598 memory appetite. When amount of memory allocated by UDP exceeds
599 this number, UDP starts to moderate memory usage.
601 pressure: This value was introduced to follow format of tcp_mem.
603 max: Number of pages allowed for queueing by all UDP sockets.
605 Default is calculated at boot time from amount of available memory.
607 udp_rmem_min - INTEGER
608 Minimal size of receive buffer used by UDP sockets in moderation.
609 Each UDP socket is able to use the size for receiving data, even if
610 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
613 udp_wmem_min - INTEGER
614 Minimal size of send buffer used by UDP sockets in moderation.
615 Each UDP socket is able to use the size for sending data, even if
616 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
621 cipso_cache_enable - BOOLEAN
622 If set, enable additions to and lookups from the CIPSO label mapping
623 cache. If unset, additions are ignored and lookups always result in a
624 miss. However, regardless of the setting the cache is still
625 invalidated when required when means you can safely toggle this on and
626 off and the cache will always be "safe".
629 cipso_cache_bucket_size - INTEGER
630 The CIPSO label cache consists of a fixed size hash table with each
631 hash bucket containing a number of cache entries. This variable limits
632 the number of entries in each hash bucket; the larger the value the
633 more CIPSO label mappings that can be cached. When the number of
634 entries in a given hash bucket reaches this limit adding new entries
635 causes the oldest entry in the bucket to be removed to make room.
638 cipso_rbm_optfmt - BOOLEAN
639 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
640 the CIPSO draft specification (see Documentation/netlabel for details).
641 This means that when set the CIPSO tag will be padded with empty
642 categories in order to make the packet data 32-bit aligned.
645 cipso_rbm_structvalid - BOOLEAN
646 If set, do a very strict check of the CIPSO option when
647 ip_options_compile() is called. If unset, relax the checks done during
648 ip_options_compile(). Either way is "safe" as errors are caught else
649 where in the CIPSO processing code but setting this to 0 (False) should
650 result in less work (i.e. it should be faster) but could cause problems
651 with other implementations that require strict checking.
656 ip_local_port_range - 2 INTEGERS
657 Defines the local port range that is used by TCP and UDP to
658 choose the local port. The first number is the first, the
659 second the last local port number. The default values are
660 32768 and 61000 respectively.
662 ip_local_reserved_ports - list of comma separated ranges
663 Specify the ports which are reserved for known third-party
664 applications. These ports will not be used by automatic port
665 assignments (e.g. when calling connect() or bind() with port
666 number 0). Explicit port allocation behavior is unchanged.
668 The format used for both input and output is a comma separated
669 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
670 10). Writing to the file will clear all previously reserved
671 ports and update the current list with the one given in the
674 Note that ip_local_port_range and ip_local_reserved_ports
675 settings are independent and both are considered by the kernel
676 when determining which ports are available for automatic port
679 You can reserve ports which are not in the current
680 ip_local_port_range, e.g.:
682 $ cat /proc/sys/net/ipv4/ip_local_port_range
684 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
687 although this is redundant. However such a setting is useful
688 if later the port range is changed to a value that will
689 include the reserved ports.
693 ip_nonlocal_bind - BOOLEAN
694 If set, allows processes to bind() to non-local IP addresses,
695 which can be quite useful - but may break some applications.
699 If set non-zero, enables support for dynamic addresses.
700 If set to a non-zero value larger than 1, a kernel log
701 message will be printed when dynamic address rewriting
705 ip_early_demux - BOOLEAN
706 Optimize input packet processing down to one demux for
707 certain kinds of local sockets. Currently we only do this
708 for established TCP sockets.
710 It may add an additional cost for pure routing workloads that
711 reduces overall throughput, in such case you should disable it.
714 icmp_echo_ignore_all - BOOLEAN
715 If set non-zero, then the kernel will ignore all ICMP ECHO
719 icmp_echo_ignore_broadcasts - BOOLEAN
720 If set non-zero, then the kernel will ignore all ICMP ECHO and
721 TIMESTAMP requests sent to it via broadcast/multicast.
724 icmp_ratelimit - INTEGER
725 Limit the maximal rates for sending ICMP packets whose type matches
726 icmp_ratemask (see below) to specific targets.
727 0 to disable any limiting,
728 otherwise the minimal space between responses in milliseconds.
731 icmp_ratemask - INTEGER
732 Mask made of ICMP types for which rates are being limited.
733 Significant bits: IHGFEDCBA9876543210
734 Default mask: 0000001100000011000 (6168)
736 Bit definitions (see include/linux/icmp.h):
738 3 Destination Unreachable *
743 C Parameter Problem *
748 H Address Mask Request
751 * These are rate limited by default (see default mask above)
753 icmp_ignore_bogus_error_responses - BOOLEAN
754 Some routers violate RFC1122 by sending bogus responses to broadcast
755 frames. Such violations are normally logged via a kernel warning.
756 If this is set to TRUE, the kernel will not give such warnings, which
757 will avoid log file clutter.
760 icmp_errors_use_inbound_ifaddr - BOOLEAN
762 If zero, icmp error messages are sent with the primary address of
763 the exiting interface.
765 If non-zero, the message will be sent with the primary address of
766 the interface that received the packet that caused the icmp error.
767 This is the behaviour network many administrators will expect from
768 a router. And it can make debugging complicated network layouts
771 Note that if no primary address exists for the interface selected,
772 then the primary address of the first non-loopback interface that
773 has one will be used regardless of this setting.
777 igmp_max_memberships - INTEGER
778 Change the maximum number of multicast groups we can subscribe to.
781 Theoretical maximum value is bounded by having to send a membership
782 report in a single datagram (i.e. the report can't span multiple
783 datagrams, or risk confusing the switch and leaving groups you don't
786 The number of supported groups 'M' is bounded by the number of group
787 report entries you can fit into a single datagram of 65535 bytes.
789 M = 65536-sizeof (ip header)/(sizeof(Group record))
791 Group records are variable length, with a minimum of 12 bytes.
792 So net.ipv4.igmp_max_memberships should not be set higher than:
794 (65536-24) / 12 = 5459
796 The value 5459 assumes no IP header options, so in practice
797 this number may be lower.
799 conf/interface/* changes special settings per interface (where
800 "interface" is the name of your network interface)
802 conf/all/* is special, changes the settings for all interfaces
804 log_martians - BOOLEAN
805 Log packets with impossible addresses to kernel log.
806 log_martians for the interface will be enabled if at least one of
807 conf/{all,interface}/log_martians is set to TRUE,
808 it will be disabled otherwise
810 accept_redirects - BOOLEAN
811 Accept ICMP redirect messages.
812 accept_redirects for the interface will be enabled if:
813 - both conf/{all,interface}/accept_redirects are TRUE in the case
814 forwarding for the interface is enabled
816 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
817 case forwarding for the interface is disabled
818 accept_redirects for the interface will be disabled otherwise
823 Enable IP forwarding on this interface.
825 mc_forwarding - BOOLEAN
826 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
827 and a multicast routing daemon is required.
828 conf/all/mc_forwarding must also be set to TRUE to enable multicast
829 routing for the interface
832 Integer value used to differentiate the devices by the medium they
833 are attached to. Two devices can have different id values when
834 the broadcast packets are received only on one of them.
835 The default value 0 means that the device is the only interface
836 to its medium, value of -1 means that medium is not known.
838 Currently, it is used to change the proxy_arp behavior:
839 the proxy_arp feature is enabled for packets forwarded between
840 two devices attached to different media.
844 proxy_arp for the interface will be enabled if at least one of
845 conf/{all,interface}/proxy_arp is set to TRUE,
846 it will be disabled otherwise
848 proxy_arp_pvlan - BOOLEAN
849 Private VLAN proxy arp.
850 Basically allow proxy arp replies back to the same interface
851 (from which the ARP request/solicitation was received).
853 This is done to support (ethernet) switch features, like RFC
854 3069, where the individual ports are NOT allowed to
855 communicate with each other, but they are allowed to talk to
856 the upstream router. As described in RFC 3069, it is possible
857 to allow these hosts to communicate through the upstream
858 router by proxy_arp'ing. Don't need to be used together with
861 This technology is known by different names:
862 In RFC 3069 it is called VLAN Aggregation.
863 Cisco and Allied Telesyn call it Private VLAN.
864 Hewlett-Packard call it Source-Port filtering or port-isolation.
865 Ericsson call it MAC-Forced Forwarding (RFC Draft).
867 shared_media - BOOLEAN
868 Send(router) or accept(host) RFC1620 shared media redirects.
869 Overrides ip_secure_redirects.
870 shared_media for the interface will be enabled if at least one of
871 conf/{all,interface}/shared_media is set to TRUE,
872 it will be disabled otherwise
875 secure_redirects - BOOLEAN
876 Accept ICMP redirect messages only for gateways,
877 listed in default gateway list.
878 secure_redirects for the interface will be enabled if at least one of
879 conf/{all,interface}/secure_redirects is set to TRUE,
880 it will be disabled otherwise
883 send_redirects - BOOLEAN
884 Send redirects, if router.
885 send_redirects for the interface will be enabled if at least one of
886 conf/{all,interface}/send_redirects is set to TRUE,
887 it will be disabled otherwise
890 bootp_relay - BOOLEAN
891 Accept packets with source address 0.b.c.d destined
892 not to this host as local ones. It is supposed, that
893 BOOTP relay daemon will catch and forward such packets.
894 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
899 accept_source_route - BOOLEAN
900 Accept packets with SRR option.
901 conf/all/accept_source_route must also be set to TRUE to accept packets
902 with SRR option on the interface
903 default TRUE (router)
906 accept_local - BOOLEAN
907 Accept packets with local source addresses. In combination
908 with suitable routing, this can be used to direct packets
909 between two local interfaces over the wire and have them
912 rp_filter must be set to a non-zero value in order for
913 accept_local to have an effect.
917 route_localnet - BOOLEAN
918 Do not consider loopback addresses as martian source or destination
919 while routing. This enables the use of 127/8 for local routing purposes.
923 0 - No source validation.
924 1 - Strict mode as defined in RFC3704 Strict Reverse Path
925 Each incoming packet is tested against the FIB and if the interface
926 is not the best reverse path the packet check will fail.
927 By default failed packets are discarded.
928 2 - Loose mode as defined in RFC3704 Loose Reverse Path
929 Each incoming packet's source address is also tested against the FIB
930 and if the source address is not reachable via any interface
931 the packet check will fail.
933 Current recommended practice in RFC3704 is to enable strict mode
934 to prevent IP spoofing from DDos attacks. If using asymmetric routing
935 or other complicated routing, then loose mode is recommended.
937 The max value from conf/{all,interface}/rp_filter is used
938 when doing source validation on the {interface}.
940 Default value is 0. Note that some distributions enable it
944 1 - Allows you to have multiple network interfaces on the same
945 subnet, and have the ARPs for each interface be answered
946 based on whether or not the kernel would route a packet from
947 the ARP'd IP out that interface (therefore you must use source
948 based routing for this to work). In other words it allows control
949 of which cards (usually 1) will respond to an arp request.
951 0 - (default) The kernel can respond to arp requests with addresses
952 from other interfaces. This may seem wrong but it usually makes
953 sense, because it increases the chance of successful communication.
954 IP addresses are owned by the complete host on Linux, not by
955 particular interfaces. Only for more complex setups like load-
956 balancing, does this behaviour cause problems.
958 arp_filter for the interface will be enabled if at least one of
959 conf/{all,interface}/arp_filter is set to TRUE,
960 it will be disabled otherwise
962 arp_announce - INTEGER
963 Define different restriction levels for announcing the local
964 source IP address from IP packets in ARP requests sent on
966 0 - (default) Use any local address, configured on any interface
967 1 - Try to avoid local addresses that are not in the target's
968 subnet for this interface. This mode is useful when target
969 hosts reachable via this interface require the source IP
970 address in ARP requests to be part of their logical network
971 configured on the receiving interface. When we generate the
972 request we will check all our subnets that include the
973 target IP and will preserve the source address if it is from
974 such subnet. If there is no such subnet we select source
975 address according to the rules for level 2.
976 2 - Always use the best local address for this target.
977 In this mode we ignore the source address in the IP packet
978 and try to select local address that we prefer for talks with
979 the target host. Such local address is selected by looking
980 for primary IP addresses on all our subnets on the outgoing
981 interface that include the target IP address. If no suitable
982 local address is found we select the first local address
983 we have on the outgoing interface or on all other interfaces,
984 with the hope we will receive reply for our request and
985 even sometimes no matter the source IP address we announce.
987 The max value from conf/{all,interface}/arp_announce is used.
989 Increasing the restriction level gives more chance for
990 receiving answer from the resolved target while decreasing
991 the level announces more valid sender's information.
994 Define different modes for sending replies in response to
995 received ARP requests that resolve local target IP addresses:
996 0 - (default): reply for any local target IP address, configured
998 1 - reply only if the target IP address is local address
999 configured on the incoming interface
1000 2 - reply only if the target IP address is local address
1001 configured on the incoming interface and both with the
1002 sender's IP address are part from same subnet on this interface
1003 3 - do not reply for local addresses configured with scope host,
1004 only resolutions for global and link addresses are replied
1006 8 - do not reply for all local addresses
1008 The max value from conf/{all,interface}/arp_ignore is used
1009 when ARP request is received on the {interface}
1011 arp_notify - BOOLEAN
1012 Define mode for notification of address and device changes.
1013 0 - (default): do nothing
1014 1 - Generate gratuitous arp requests when device is brought up
1015 or hardware address changes.
1017 arp_accept - BOOLEAN
1018 Define behavior for gratuitous ARP frames who's IP is not
1019 already present in the ARP table:
1020 0 - don't create new entries in the ARP table
1021 1 - create new entries in the ARP table
1023 Both replies and requests type gratuitous arp will trigger the
1024 ARP table to be updated, if this setting is on.
1026 If the ARP table already contains the IP address of the
1027 gratuitous arp frame, the arp table will be updated regardless
1028 if this setting is on or off.
1031 app_solicit - INTEGER
1032 The maximum number of probes to send to the user space ARP daemon
1033 via netlink before dropping back to multicast probes (see
1034 mcast_solicit). Defaults to 0.
1036 disable_policy - BOOLEAN
1037 Disable IPSEC policy (SPD) for this interface
1039 disable_xfrm - BOOLEAN
1040 Disable IPSEC encryption on this interface, whatever the policy
1045 Allows you to write a number, which can be used as required.
1049 kuznet@ms2.inr.ac.ru
1055 delon.nicolas@wanadoo.fr
1060 /proc/sys/net/ipv6/* Variables:
1062 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1063 apply to IPv6 [XXX?].
1065 bindv6only - BOOLEAN
1066 Default value for IPV6_V6ONLY socket option,
1067 which restricts use of the IPv6 socket to IPv6 communication
1069 TRUE: disable IPv4-mapped address feature
1070 FALSE: enable IPv4-mapped address feature
1072 Default: FALSE (as specified in RFC3493)
1076 ip6frag_high_thresh - INTEGER
1077 Maximum memory used to reassemble IPv6 fragments. When
1078 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1079 the fragment handler will toss packets until ip6frag_low_thresh
1082 ip6frag_low_thresh - INTEGER
1083 See ip6frag_high_thresh
1085 ip6frag_time - INTEGER
1086 Time in seconds to keep an IPv6 fragment in memory.
1088 ip6frag_secret_interval - INTEGER
1089 Regeneration interval (in seconds) of the hash secret (or lifetime
1090 for the hash secret) for IPv6 fragments.
1094 Change the interface-specific default settings.
1098 Change all the interface-specific settings.
1100 [XXX: Other special features than forwarding?]
1102 conf/all/forwarding - BOOLEAN
1103 Enable global IPv6 forwarding between all interfaces.
1105 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1106 to control which interfaces may forward packets and which not.
1108 This also sets all interfaces' Host/Router setting
1109 'forwarding' to the specified value. See below for details.
1111 This referred to as global forwarding.
1117 Change special settings per interface.
1119 The functional behaviour for certain settings is different
1120 depending on whether local forwarding is enabled or not.
1123 Accept Router Advertisements; autoconfigure using them.
1125 It also determines whether or not to transmit Router
1126 Solicitations. If and only if the functional setting is to
1127 accept Router Advertisements, Router Solicitations will be
1130 Possible values are:
1131 0 Do not accept Router Advertisements.
1132 1 Accept Router Advertisements if forwarding is disabled.
1133 2 Overrule forwarding behaviour. Accept Router Advertisements
1134 even if forwarding is enabled.
1136 Functional default: enabled if local forwarding is disabled.
1137 disabled if local forwarding is enabled.
1139 accept_ra_defrtr - BOOLEAN
1140 Learn default router in Router Advertisement.
1142 Functional default: enabled if accept_ra is enabled.
1143 disabled if accept_ra is disabled.
1145 accept_ra_pinfo - BOOLEAN
1146 Learn Prefix Information in Router Advertisement.
1148 Functional default: enabled if accept_ra is enabled.
1149 disabled if accept_ra is disabled.
1151 accept_ra_rt_info_max_plen - INTEGER
1152 Maximum prefix length of Route Information in RA.
1154 Route Information w/ prefix larger than or equal to this
1155 variable shall be ignored.
1157 Functional default: 0 if accept_ra_rtr_pref is enabled.
1158 -1 if accept_ra_rtr_pref is disabled.
1160 accept_ra_rtr_pref - BOOLEAN
1161 Accept Router Preference in RA.
1163 Functional default: enabled if accept_ra is enabled.
1164 disabled if accept_ra is disabled.
1166 accept_redirects - BOOLEAN
1169 Functional default: enabled if local forwarding is disabled.
1170 disabled if local forwarding is enabled.
1172 accept_source_route - INTEGER
1173 Accept source routing (routing extension header).
1175 >= 0: Accept only routing header type 2.
1176 < 0: Do not accept routing header.
1181 Autoconfigure addresses using Prefix Information in Router
1184 Functional default: enabled if accept_ra_pinfo is enabled.
1185 disabled if accept_ra_pinfo is disabled.
1187 dad_transmits - INTEGER
1188 The amount of Duplicate Address Detection probes to send.
1191 forwarding - INTEGER
1192 Configure interface-specific Host/Router behaviour.
1194 Note: It is recommended to have the same setting on all
1195 interfaces; mixed router/host scenarios are rather uncommon.
1197 Possible values are:
1198 0 Forwarding disabled
1199 1 Forwarding enabled
1203 By default, Host behaviour is assumed. This means:
1205 1. IsRouter flag is not set in Neighbour Advertisements.
1206 2. If accept_ra is TRUE (default), transmit Router
1208 3. If accept_ra is TRUE (default), accept Router
1209 Advertisements (and do autoconfiguration).
1210 4. If accept_redirects is TRUE (default), accept Redirects.
1214 If local forwarding is enabled, Router behaviour is assumed.
1215 This means exactly the reverse from the above:
1217 1. IsRouter flag is set in Neighbour Advertisements.
1218 2. Router Solicitations are not sent unless accept_ra is 2.
1219 3. Router Advertisements are ignored unless accept_ra is 2.
1220 4. Redirects are ignored.
1222 Default: 0 (disabled) if global forwarding is disabled (default),
1223 otherwise 1 (enabled).
1226 Default Hop Limit to set.
1230 Default Maximum Transfer Unit
1231 Default: 1280 (IPv6 required minimum)
1233 router_probe_interval - INTEGER
1234 Minimum interval (in seconds) between Router Probing described
1239 router_solicitation_delay - INTEGER
1240 Number of seconds to wait after interface is brought up
1241 before sending Router Solicitations.
1244 router_solicitation_interval - INTEGER
1245 Number of seconds to wait between Router Solicitations.
1248 router_solicitations - INTEGER
1249 Number of Router Solicitations to send until assuming no
1250 routers are present.
1253 use_tempaddr - INTEGER
1254 Preference for Privacy Extensions (RFC3041).
1255 <= 0 : disable Privacy Extensions
1256 == 1 : enable Privacy Extensions, but prefer public
1257 addresses over temporary addresses.
1258 > 1 : enable Privacy Extensions and prefer temporary
1259 addresses over public addresses.
1260 Default: 0 (for most devices)
1261 -1 (for point-to-point devices and loopback devices)
1263 temp_valid_lft - INTEGER
1264 valid lifetime (in seconds) for temporary addresses.
1265 Default: 604800 (7 days)
1267 temp_prefered_lft - INTEGER
1268 Preferred lifetime (in seconds) for temporary addresses.
1269 Default: 86400 (1 day)
1271 max_desync_factor - INTEGER
1272 Maximum value for DESYNC_FACTOR, which is a random value
1273 that ensures that clients don't synchronize with each
1274 other and generate new addresses at exactly the same time.
1275 value is in seconds.
1278 regen_max_retry - INTEGER
1279 Number of attempts before give up attempting to generate
1280 valid temporary addresses.
1283 max_addresses - INTEGER
1284 Maximum number of autoconfigured addresses per interface. Setting
1285 to zero disables the limitation. It is not recommended to set this
1286 value too large (or to zero) because it would be an easy way to
1287 crash the kernel by allowing too many addresses to be created.
1290 disable_ipv6 - BOOLEAN
1291 Disable IPv6 operation. If accept_dad is set to 2, this value
1292 will be dynamically set to TRUE if DAD fails for the link-local
1294 Default: FALSE (enable IPv6 operation)
1296 When this value is changed from 1 to 0 (IPv6 is being enabled),
1297 it will dynamically create a link-local address on the given
1298 interface and start Duplicate Address Detection, if necessary.
1300 When this value is changed from 0 to 1 (IPv6 is being disabled),
1301 it will dynamically delete all address on the given interface.
1303 accept_dad - INTEGER
1304 Whether to accept DAD (Duplicate Address Detection).
1306 1: Enable DAD (default)
1307 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1308 link-local address has been found.
1310 force_tllao - BOOLEAN
1311 Enable sending the target link-layer address option even when
1312 responding to a unicast neighbor solicitation.
1315 Quoting from RFC 2461, section 4.4, Target link-layer address:
1317 "The option MUST be included for multicast solicitations in order to
1318 avoid infinite Neighbor Solicitation "recursion" when the peer node
1319 does not have a cache entry to return a Neighbor Advertisements
1320 message. When responding to unicast solicitations, the option can be
1321 omitted since the sender of the solicitation has the correct link-
1322 layer address; otherwise it would not have be able to send the unicast
1323 solicitation in the first place. However, including the link-layer
1324 address in this case adds little overhead and eliminates a potential
1325 race condition where the sender deletes the cached link-layer address
1326 prior to receiving a response to a previous solicitation."
1328 ndisc_notify - BOOLEAN
1329 Define mode for notification of address and device changes.
1330 0 - (default): do nothing
1331 1 - Generate unsolicited neighbour advertisements when device is brought
1332 up or hardware address changes.
1336 Limit the maximal rates for sending ICMPv6 packets.
1337 0 to disable any limiting,
1338 otherwise the minimal space between responses in milliseconds.
1343 Pekka Savola <pekkas@netcore.fi>
1344 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1347 /proc/sys/net/bridge/* Variables:
1349 bridge-nf-call-arptables - BOOLEAN
1350 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1354 bridge-nf-call-iptables - BOOLEAN
1355 1 : pass bridged IPv4 traffic to iptables' chains.
1359 bridge-nf-call-ip6tables - BOOLEAN
1360 1 : pass bridged IPv6 traffic to ip6tables' chains.
1364 bridge-nf-filter-vlan-tagged - BOOLEAN
1365 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1369 bridge-nf-filter-pppoe-tagged - BOOLEAN
1370 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1374 bridge-nf-pass-vlan-input-dev - BOOLEAN
1375 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1376 interface on the bridge and set the netfilter input device to the vlan.
1377 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1378 target work with vlan-on-top-of-bridge interfaces. When no matching
1379 vlan interface is found, or this switch is off, the input device is
1380 set to the bridge interface.
1381 0: disable bridge netfilter vlan interface lookup.
1384 proc/sys/net/sctp/* Variables:
1386 addip_enable - BOOLEAN
1387 Enable or disable extension of Dynamic Address Reconfiguration
1388 (ADD-IP) functionality specified in RFC5061. This extension provides
1389 the ability to dynamically add and remove new addresses for the SCTP
1392 1: Enable extension.
1394 0: Disable extension.
1398 addip_noauth_enable - BOOLEAN
1399 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1400 authentication to protect the operations of adding or removing new
1401 addresses. This requirement is mandated so that unauthorized hosts
1402 would not be able to hijack associations. However, older
1403 implementations may not have implemented this requirement while
1404 allowing the ADD-IP extension. For reasons of interoperability,
1405 we provide this variable to control the enforcement of the
1406 authentication requirement.
1408 1: Allow ADD-IP extension to be used without authentication. This
1409 should only be set in a closed environment for interoperability
1410 with older implementations.
1412 0: Enforce the authentication requirement
1416 auth_enable - BOOLEAN
1417 Enable or disable Authenticated Chunks extension. This extension
1418 provides the ability to send and receive authenticated chunks and is
1419 required for secure operation of Dynamic Address Reconfiguration
1422 1: Enable this extension.
1423 0: Disable this extension.
1427 prsctp_enable - BOOLEAN
1428 Enable or disable the Partial Reliability extension (RFC3758) which
1429 is used to notify peers that a given DATA should no longer be expected.
1437 The limit of the number of new packets that can be initially sent. It
1438 controls how bursty the generated traffic can be.
1442 association_max_retrans - INTEGER
1443 Set the maximum number for retransmissions that an association can
1444 attempt deciding that the remote end is unreachable. If this value
1445 is exceeded, the association is terminated.
1449 max_init_retransmits - INTEGER
1450 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1451 that an association will attempt before declaring the destination
1452 unreachable and terminating.
1456 path_max_retrans - INTEGER
1457 The maximum number of retransmissions that will be attempted on a given
1458 path. Once this threshold is exceeded, the path is considered
1459 unreachable, and new traffic will use a different path when the
1460 association is multihomed.
1464 pf_retrans - INTEGER
1465 The number of retransmissions that will be attempted on a given path
1466 before traffic is redirected to an alternate transport (should one
1467 exist). Note this is distinct from path_max_retrans, as a path that
1468 passes the pf_retrans threshold can still be used. Its only
1469 deprioritized when a transmission path is selected by the stack. This
1470 setting is primarily used to enable fast failover mechanisms without
1471 having to reduce path_max_retrans to a very low value. See:
1472 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1473 for details. Note also that a value of pf_retrans > path_max_retrans
1474 disables this feature
1478 rto_initial - INTEGER
1479 The initial round trip timeout value in milliseconds that will be used
1480 in calculating round trip times. This is the initial time interval
1481 for retransmissions.
1486 The maximum value (in milliseconds) of the round trip timeout. This
1487 is the largest time interval that can elapse between retransmissions.
1492 The minimum value (in milliseconds) of the round trip timeout. This
1493 is the smallest time interval the can elapse between retransmissions.
1497 hb_interval - INTEGER
1498 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1499 are sent at the specified interval on idle paths to probe the state of
1500 a given path between 2 associations.
1504 sack_timeout - INTEGER
1505 The amount of time (in milliseconds) that the implementation will wait
1510 checksum_disable - BOOLEAN
1511 Disable SCTP checksum computing and verification for debugging purpose.
1513 1: Disable checksumming
1514 0: Enable checksumming
1518 valid_cookie_life - INTEGER
1519 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1520 is used during association establishment.
1524 cookie_preserve_enable - BOOLEAN
1525 Enable or disable the ability to extend the lifetime of the SCTP cookie
1526 that is used during the establishment phase of SCTP association
1528 1: Enable cookie lifetime extension.
1533 cookie_hmac_alg - STRING
1534 Select the hmac algorithm used when generating the cookie value sent by
1535 a listening sctp socket to a connecting client in the INIT-ACK chunk.
1540 Ability to assign md5 or sha1 as the selected alg is predicated on the
1541 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
1542 CONFIG_CRYPTO_SHA1).
1544 Default: Dependent on configuration. MD5 if available, else SHA1 if
1545 available, else none.
1547 rcvbuf_policy - INTEGER
1548 Determines if the receive buffer is attributed to the socket or to
1549 association. SCTP supports the capability to create multiple
1550 associations on a single socket. When using this capability, it is
1551 possible that a single stalled association that's buffering a lot
1552 of data may block other associations from delivering their data by
1553 consuming all of the receive buffer space. To work around this,
1554 the rcvbuf_policy could be set to attribute the receiver buffer space
1555 to each association instead of the socket. This prevents the described
1558 1: rcvbuf space is per association
1559 0: rcvbuf space is per socket
1563 sndbuf_policy - INTEGER
1564 Similar to rcvbuf_policy above, this applies to send buffer space.
1566 1: Send buffer is tracked per association
1567 0: Send buffer is tracked per socket.
1571 sctp_mem - vector of 3 INTEGERs: min, pressure, max
1572 Number of pages allowed for queueing by all SCTP sockets.
1574 min: Below this number of pages SCTP is not bothered about its
1575 memory appetite. When amount of memory allocated by SCTP exceeds
1576 this number, SCTP starts to moderate memory usage.
1578 pressure: This value was introduced to follow format of tcp_mem.
1580 max: Number of pages allowed for queueing by all SCTP sockets.
1582 Default is calculated at boot time from amount of available memory.
1584 sctp_rmem - vector of 3 INTEGERs: min, default, max
1585 Only the first value ("min") is used, "default" and "max" are
1588 min: Minimal size of receive buffer used by SCTP socket.
1589 It is guaranteed to each SCTP socket (but not association) even
1590 under moderate memory pressure.
1594 sctp_wmem - vector of 3 INTEGERs: min, default, max
1595 Currently this tunable has no effect.
1597 addr_scope_policy - INTEGER
1598 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1600 0 - Disable IPv4 address scoping
1601 1 - Enable IPv4 address scoping
1602 2 - Follow draft but allow IPv4 private addresses
1603 3 - Follow draft but allow IPv4 link local addresses
1608 /proc/sys/net/core/*
1609 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
1612 /proc/sys/net/unix/*
1613 max_dgram_qlen - INTEGER
1614 The maximum length of dgram socket receive queue
1621 /proc/sys/net/irda/*
1622 fast_poll_increase FIXME
1623 warn_noreply_time FIXME
1624 discovery_slots FIXME
1627 discovery_timeout FIXME
1628 lap_keepalive_time FIXME
1629 max_noreply_time FIXME
1630 max_tx_data_size FIXME
1632 min_tx_turn_time FIXME