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
16 ip_no_pmtu_disc - BOOLEAN
17 Disable Path MTU Discovery.
21 default 562 - minimum discovered Path MTU
24 Time, in seconds, that cached PMTU information is kept.
27 The advertised MSS depends on the first hop route MTU, but will
28 never be lower than this setting.
30 rt_cache_rebuild_count - INTEGER
31 The per net-namespace route cache emergency rebuild threshold.
32 Any net-namespace having its route cache rebuilt due to
33 a hash bucket chain being too long more than this many times
34 will have its route caching disabled
38 ipfrag_high_thresh - INTEGER
39 Maximum memory used to reassemble IP fragments. When
40 ipfrag_high_thresh bytes of memory is allocated for this purpose,
41 the fragment handler will toss packets until ipfrag_low_thresh
44 ipfrag_low_thresh - INTEGER
45 See ipfrag_high_thresh
48 Time in seconds to keep an IP fragment in memory.
50 ipfrag_secret_interval - INTEGER
51 Regeneration interval (in seconds) of the hash secret (or lifetime
52 for the hash secret) for IP fragments.
55 ipfrag_max_dist - INTEGER
56 ipfrag_max_dist is a non-negative integer value which defines the
57 maximum "disorder" which is allowed among fragments which share a
58 common IP source address. Note that reordering of packets is
59 not unusual, but if a large number of fragments arrive from a source
60 IP address while a particular fragment queue remains incomplete, it
61 probably indicates that one or more fragments belonging to that queue
62 have been lost. When ipfrag_max_dist is positive, an additional check
63 is done on fragments before they are added to a reassembly queue - if
64 ipfrag_max_dist (or more) fragments have arrived from a particular IP
65 address between additions to any IP fragment queue using that source
66 address, it's presumed that one or more fragments in the queue are
67 lost. The existing fragment queue will be dropped, and a new one
68 started. An ipfrag_max_dist value of zero disables this check.
70 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
71 result in unnecessarily dropping fragment queues when normal
72 reordering of packets occurs, which could lead to poor application
73 performance. Using a very large value, e.g. 50000, increases the
74 likelihood of incorrectly reassembling IP fragments that originate
75 from different IP datagrams, which could result in data corruption.
80 inet_peer_threshold - INTEGER
81 The approximate size of the storage. Starting from this threshold
82 entries will be thrown aggressively. This threshold also determines
83 entries' time-to-live and time intervals between garbage collection
84 passes. More entries, less time-to-live, less GC interval.
86 inet_peer_minttl - INTEGER
87 Minimum time-to-live of entries. Should be enough to cover fragment
88 time-to-live on the reassembling side. This minimum time-to-live is
89 guaranteed if the pool size is less than inet_peer_threshold.
92 inet_peer_maxttl - INTEGER
93 Maximum time-to-live of entries. Unused entries will expire after
94 this period of time if there is no memory pressure on the pool (i.e.
95 when the number of entries in the pool is very small).
98 inet_peer_gc_mintime - INTEGER
99 Minimum interval between garbage collection passes. This interval is
100 in effect under high memory pressure on the pool.
103 inet_peer_gc_maxtime - INTEGER
104 Minimum interval between garbage collection passes. This interval is
105 in effect under low (or absent) memory pressure on the pool.
111 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
112 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
116 Controls Appropriate Byte Count (ABC) defined in RFC3465.
117 ABC is a way of increasing congestion window (cwnd) more slowly
118 in response to partial acknowledgments.
120 0 increase cwnd once per acknowledgment (no ABC)
121 1 increase cwnd once per acknowledgment of full sized segment
122 2 allow increase cwnd by two if acknowledgment is
123 of two segments to compensate for delayed acknowledgments.
126 tcp_abort_on_overflow - BOOLEAN
127 If listening service is too slow to accept new connections,
128 reset them. Default state is FALSE. It means that if overflow
129 occurred due to a burst, connection will recover. Enable this
130 option _only_ if you are really sure that listening daemon
131 cannot be tuned to accept connections faster. Enabling this
132 option can harm clients of your server.
134 tcp_adv_win_scale - INTEGER
135 Count buffering overhead as bytes/2^tcp_adv_win_scale
136 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
140 tcp_allowed_congestion_control - STRING
141 Show/set the congestion control choices available to non-privileged
142 processes. The list is a subset of those listed in
143 tcp_available_congestion_control.
144 Default is "reno" and the default setting (tcp_congestion_control).
146 tcp_app_win - INTEGER
147 Reserve max(window/2^tcp_app_win, mss) of window for application
148 buffer. Value 0 is special, it means that nothing is reserved.
151 tcp_available_congestion_control - STRING
152 Shows the available congestion control choices that are registered.
153 More congestion control algorithms may be available as modules,
156 tcp_base_mss - INTEGER
157 The initial value of search_low to be used by the packetization layer
158 Path MTU discovery (MTU probing). If MTU probing is enabled,
159 this is the initial MSS used by the connection.
161 tcp_congestion_control - STRING
162 Set the congestion control algorithm to be used for new
163 connections. The algorithm "reno" is always available, but
164 additional choices may be available based on kernel configuration.
165 Default is set as part of kernel configuration.
167 tcp_cookie_size - INTEGER
168 Default size of TCP Cookie Transactions (TCPCT) option, that may be
169 overridden on a per socket basis by the TCPCT socket option.
170 Values greater than the maximum (16) are interpreted as the maximum.
171 Values greater than zero and less than the minimum (8) are interpreted
172 as the minimum. Odd values are interpreted as the next even value.
176 Allows TCP to send "duplicate" SACKs.
179 Enable Explicit Congestion Notification (ECN) in TCP. ECN is only
180 used when both ends of the TCP flow support it. It is useful to
181 avoid losses due to congestion (when the bottleneck router supports
186 2 Only server-side ECN enabled. If the other end does
187 not support ECN, behavior is like with ECN disabled.
191 Enable FACK congestion avoidance and fast retransmission.
192 The value is not used, if tcp_sack is not enabled.
194 tcp_fin_timeout - INTEGER
195 Time to hold socket in state FIN-WAIT-2, if it was closed
196 by our side. Peer can be broken and never close its side,
197 or even died unexpectedly. Default value is 60sec.
198 Usual value used in 2.2 was 180 seconds, you may restore
199 it, but remember that if your machine is even underloaded WEB server,
200 you risk to overflow memory with kilotons of dead sockets,
201 FIN-WAIT-2 sockets are less dangerous than FIN-WAIT-1,
202 because they eat maximum 1.5K of memory, but they tend
203 to live longer. Cf. tcp_max_orphans.
206 Enables Forward RTO-Recovery (F-RTO) defined in RFC4138.
207 F-RTO is an enhanced recovery algorithm for TCP retransmission
208 timeouts. It is particularly beneficial in wireless environments
209 where packet loss is typically due to random radio interference
210 rather than intermediate router congestion. F-RTO is sender-side
211 only modification. Therefore it does not require any support from
214 If set to 1, basic version is enabled. 2 enables SACK enhanced
215 F-RTO if flow uses SACK. The basic version can be used also when
216 SACK is in use though scenario(s) with it exists where F-RTO
217 interacts badly with the packet counting of the SACK enabled TCP
220 tcp_frto_response - INTEGER
221 When F-RTO has detected that a TCP retransmission timeout was
222 spurious (i.e, the timeout would have been avoided had TCP set a
223 longer retransmission timeout), TCP has several options what to do
224 next. Possible values are:
225 0 Rate halving based; a smooth and conservative response,
226 results in halved cwnd and ssthresh after one RTT
227 1 Very conservative response; not recommended because even
228 though being valid, it interacts poorly with the rest of
229 Linux TCP, halves cwnd and ssthresh immediately
230 2 Aggressive response; undoes congestion control measures
231 that are now known to be unnecessary (ignoring the
232 possibility of a lost retransmission that would require
233 TCP to be more cautious), cwnd and ssthresh are restored
234 to the values prior timeout
235 Default: 0 (rate halving based)
237 tcp_keepalive_time - INTEGER
238 How often TCP sends out keepalive messages when keepalive is enabled.
241 tcp_keepalive_probes - INTEGER
242 How many keepalive probes TCP sends out, until it decides that the
243 connection is broken. Default value: 9.
245 tcp_keepalive_intvl - INTEGER
246 How frequently the probes are send out. Multiplied by
247 tcp_keepalive_probes it is time to kill not responding connection,
248 after probes started. Default value: 75sec i.e. connection
249 will be aborted after ~11 minutes of retries.
251 tcp_low_latency - BOOLEAN
252 If set, the TCP stack makes decisions that prefer lower
253 latency as opposed to higher throughput. By default, this
254 option is not set meaning that higher throughput is preferred.
255 An example of an application where this default should be
256 changed would be a Beowulf compute cluster.
259 tcp_max_orphans - INTEGER
260 Maximal number of TCP sockets not attached to any user file handle,
261 held by system. If this number is exceeded orphaned connections are
262 reset immediately and warning is printed. This limit exists
263 only to prevent simple DoS attacks, you _must_ not rely on this
264 or lower the limit artificially, but rather increase it
265 (probably, after increasing installed memory),
266 if network conditions require more than default value,
267 and tune network services to linger and kill such states
268 more aggressively. Let me to remind again: each orphan eats
269 up to ~64K of unswappable memory.
271 tcp_max_syn_backlog - INTEGER
272 Maximal number of remembered connection requests, which are
273 still did not receive an acknowledgment from connecting client.
274 Default value is 1024 for systems with more than 128Mb of memory,
275 and 128 for low memory machines. If server suffers of overload,
276 try to increase this number.
278 tcp_max_tw_buckets - INTEGER
279 Maximal number of timewait sockets held by system simultaneously.
280 If this number is exceeded time-wait socket is immediately destroyed
281 and warning is printed. This limit exists only to prevent
282 simple DoS attacks, you _must_ not lower the limit artificially,
283 but rather increase it (probably, after increasing installed memory),
284 if network conditions require more than default value.
286 tcp_mem - vector of 3 INTEGERs: min, pressure, max
287 min: below this number of pages TCP is not bothered about its
290 pressure: when amount of memory allocated by TCP exceeds this number
291 of pages, TCP moderates its memory consumption and enters memory
292 pressure mode, which is exited when memory consumption falls
295 max: number of pages allowed for queueing by all TCP sockets.
297 Defaults are calculated at boot time from amount of available
300 tcp_moderate_rcvbuf - BOOLEAN
301 If set, TCP performs receive buffer auto-tuning, attempting to
302 automatically size the buffer (no greater than tcp_rmem[2]) to
303 match the size required by the path for full throughput. Enabled by
306 tcp_mtu_probing - INTEGER
307 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
310 1 - Disabled by default, enabled when an ICMP black hole detected
311 2 - Always enabled, use initial MSS of tcp_base_mss.
313 tcp_no_metrics_save - BOOLEAN
314 By default, TCP saves various connection metrics in the route cache
315 when the connection closes, so that connections established in the
316 near future can use these to set initial conditions. Usually, this
317 increases overall performance, but may sometimes cause performance
318 degradation. If set, TCP will not cache metrics on closing
321 tcp_orphan_retries - INTEGER
322 This value influences the timeout of a locally closed TCP connection,
323 when RTO retransmissions remain unacknowledged.
324 See tcp_retries2 for more details.
326 The default value is 7.
327 If your machine is a loaded WEB server,
328 you should think about lowering this value, such sockets
329 may consume significant resources. Cf. tcp_max_orphans.
331 tcp_reordering - INTEGER
332 Maximal reordering of packets in a TCP stream.
335 tcp_retrans_collapse - BOOLEAN
336 Bug-to-bug compatibility with some broken printers.
337 On retransmit try to send bigger packets to work around bugs in
340 tcp_retries1 - INTEGER
341 This value influences the time, after which TCP decides, that
342 something is wrong due to unacknowledged RTO retransmissions,
343 and reports this suspicion to the network layer.
344 See tcp_retries2 for more details.
346 RFC 1122 recommends at least 3 retransmissions, which is the
349 tcp_retries2 - INTEGER
350 This value influences the timeout of an alive TCP connection,
351 when RTO retransmissions remain unacknowledged.
352 Given a value of N, a hypothetical TCP connection following
353 exponential backoff with an initial RTO of TCP_RTO_MIN would
354 retransmit N times before killing the connection at the (N+1)th RTO.
356 The default value of 15 yields a hypothetical timeout of 924.6
357 seconds and is a lower bound for the effective timeout.
358 TCP will effectively time out at the first RTO which exceeds the
359 hypothetical timeout.
361 RFC 1122 recommends at least 100 seconds for the timeout,
362 which corresponds to a value of at least 8.
364 tcp_rfc1337 - BOOLEAN
365 If set, the TCP stack behaves conforming to RFC1337. If unset,
366 we are not conforming to RFC, but prevent TCP TIME_WAIT
370 tcp_rmem - vector of 3 INTEGERs: min, default, max
371 min: Minimal size of receive buffer used by TCP sockets.
372 It is guaranteed to each TCP socket, even under moderate memory
376 default: initial size of receive buffer used by TCP sockets.
377 This value overrides net.core.rmem_default used by other protocols.
378 Default: 87380 bytes. This value results in window of 65535 with
379 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
380 less for default tcp_app_win. See below about these variables.
382 max: maximal size of receive buffer allowed for automatically
383 selected receiver buffers for TCP socket. This value does not override
384 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
385 automatic tuning of that socket's receive buffer size, in which
386 case this value is ignored.
387 Default: between 87380B and 4MB, depending on RAM size.
390 Enable select acknowledgments (SACKS).
392 tcp_slow_start_after_idle - BOOLEAN
393 If set, provide RFC2861 behavior and time out the congestion
394 window after an idle period. An idle period is defined at
395 the current RTO. If unset, the congestion window will not
396 be timed out after an idle period.
400 Use the Host requirements interpretation of the TCP urgent pointer field.
401 Most hosts use the older BSD interpretation, so if you turn this on
402 Linux might not communicate correctly with them.
405 tcp_synack_retries - INTEGER
406 Number of times SYNACKs for a passive TCP connection attempt will
407 be retransmitted. Should not be higher than 255. Default value
408 is 5, which corresponds to ~180seconds.
410 tcp_syncookies - BOOLEAN
411 Only valid when the kernel was compiled with CONFIG_SYNCOOKIES
412 Send out syncookies when the syn backlog queue of a socket
413 overflows. This is to prevent against the common 'SYN flood attack'
416 Note, that syncookies is fallback facility.
417 It MUST NOT be used to help highly loaded servers to stand
418 against legal connection rate. If you see SYN flood warnings
419 in your logs, but investigation shows that they occur
420 because of overload with legal connections, you should tune
421 another parameters until this warning disappear.
422 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
424 syncookies seriously violate TCP protocol, do not allow
425 to use TCP extensions, can result in serious degradation
426 of some services (f.e. SMTP relaying), visible not by you,
427 but your clients and relays, contacting you. While you see
428 SYN flood warnings in logs not being really flooded, your server
429 is seriously misconfigured.
431 tcp_syn_retries - INTEGER
432 Number of times initial SYNs for an active TCP connection attempt
433 will be retransmitted. Should not be higher than 255. Default value
434 is 5, which corresponds to ~180seconds.
436 tcp_timestamps - BOOLEAN
437 Enable timestamps as defined in RFC1323.
439 tcp_tso_win_divisor - INTEGER
440 This allows control over what percentage of the congestion window
441 can be consumed by a single TSO frame.
442 The setting of this parameter is a choice between burstiness and
443 building larger TSO frames.
446 tcp_tw_recycle - BOOLEAN
447 Enable fast recycling TIME-WAIT sockets. Default value is 0.
448 It should not be changed without advice/request of technical
451 tcp_tw_reuse - BOOLEAN
452 Allow to reuse TIME-WAIT sockets for new connections when it is
453 safe from protocol viewpoint. Default value is 0.
454 It should not be changed without advice/request of technical
457 tcp_window_scaling - BOOLEAN
458 Enable window scaling as defined in RFC1323.
460 tcp_wmem - vector of 3 INTEGERs: min, default, max
461 min: Amount of memory reserved for send buffers for TCP sockets.
462 Each TCP socket has rights to use it due to fact of its birth.
465 default: initial size of send buffer used by TCP sockets. This
466 value overrides net.core.wmem_default used by other protocols.
467 It is usually lower than net.core.wmem_default.
470 max: Maximal amount of memory allowed for automatically tuned
471 send buffers for TCP sockets. This value does not override
472 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
473 automatic tuning of that socket's send buffer size, in which case
474 this value is ignored.
475 Default: between 64K and 4MB, depending on RAM size.
477 tcp_workaround_signed_windows - BOOLEAN
478 If set, assume no receipt of a window scaling option means the
479 remote TCP is broken and treats the window as a signed quantity.
480 If unset, assume the remote TCP is not broken even if we do
481 not receive a window scaling option from them.
484 tcp_dma_copybreak - INTEGER
485 Lower limit, in bytes, of the size of socket reads that will be
486 offloaded to a DMA copy engine, if one is present in the system
487 and CONFIG_NET_DMA is enabled.
490 tcp_thin_linear_timeouts - BOOLEAN
491 Enable dynamic triggering of linear timeouts for thin streams.
492 If set, a check is performed upon retransmission by timeout to
493 determine if the stream is thin (less than 4 packets in flight).
494 As long as the stream is found to be thin, up to 6 linear
495 timeouts may be performed before exponential backoff mode is
496 initiated. This improves retransmission latency for
497 non-aggressive thin streams, often found to be time-dependent.
498 For more information on thin streams, see
499 Documentation/networking/tcp-thin.txt
502 tcp_thin_dupack - BOOLEAN
503 Enable dynamic triggering of retransmissions after one dupACK
504 for thin streams. If set, a check is performed upon reception
505 of a dupACK to determine if the stream is thin (less than 4
506 packets in flight). As long as the stream is found to be thin,
507 data is retransmitted on the first received dupACK. This
508 improves retransmission latency for non-aggressive thin
509 streams, often found to be time-dependent.
510 For more information on thin streams, see
511 Documentation/networking/tcp-thin.txt
516 udp_mem - vector of 3 INTEGERs: min, pressure, max
517 Number of pages allowed for queueing by all UDP sockets.
519 min: Below this number of pages UDP is not bothered about its
520 memory appetite. When amount of memory allocated by UDP exceeds
521 this number, UDP starts to moderate memory usage.
523 pressure: This value was introduced to follow format of tcp_mem.
525 max: Number of pages allowed for queueing by all UDP sockets.
527 Default is calculated at boot time from amount of available memory.
529 udp_rmem_min - INTEGER
530 Minimal size of receive buffer used by UDP sockets in moderation.
531 Each UDP socket is able to use the size for receiving data, even if
532 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
535 udp_wmem_min - INTEGER
536 Minimal size of send buffer used by UDP sockets in moderation.
537 Each UDP socket is able to use the size for sending data, even if
538 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
543 cipso_cache_enable - BOOLEAN
544 If set, enable additions to and lookups from the CIPSO label mapping
545 cache. If unset, additions are ignored and lookups always result in a
546 miss. However, regardless of the setting the cache is still
547 invalidated when required when means you can safely toggle this on and
548 off and the cache will always be "safe".
551 cipso_cache_bucket_size - INTEGER
552 The CIPSO label cache consists of a fixed size hash table with each
553 hash bucket containing a number of cache entries. This variable limits
554 the number of entries in each hash bucket; the larger the value the
555 more CIPSO label mappings that can be cached. When the number of
556 entries in a given hash bucket reaches this limit adding new entries
557 causes the oldest entry in the bucket to be removed to make room.
560 cipso_rbm_optfmt - BOOLEAN
561 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
562 the CIPSO draft specification (see Documentation/netlabel for details).
563 This means that when set the CIPSO tag will be padded with empty
564 categories in order to make the packet data 32-bit aligned.
567 cipso_rbm_structvalid - BOOLEAN
568 If set, do a very strict check of the CIPSO option when
569 ip_options_compile() is called. If unset, relax the checks done during
570 ip_options_compile(). Either way is "safe" as errors are caught else
571 where in the CIPSO processing code but setting this to 0 (False) should
572 result in less work (i.e. it should be faster) but could cause problems
573 with other implementations that require strict checking.
578 ip_local_port_range - 2 INTEGERS
579 Defines the local port range that is used by TCP and UDP to
580 choose the local port. The first number is the first, the
581 second the last local port number. Default value depends on
582 amount of memory available on the system:
584 < 128Mb 1024-4999 or even less.
585 This number defines number of active connections, which this
586 system can issue simultaneously to systems not supporting
587 TCP extensions (timestamps). With tcp_tw_recycle enabled
588 (i.e. by default) range 1024-4999 is enough to issue up to
589 2000 connections per second to systems supporting timestamps.
591 ip_local_reserved_ports - list of comma separated ranges
592 Specify the ports which are reserved for known third-party
593 applications. These ports will not be used by automatic port
594 assignments (e.g. when calling connect() or bind() with port
595 number 0). Explicit port allocation behavior is unchanged.
597 The format used for both input and output is a comma separated
598 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
599 10). Writing to the file will clear all previously reserved
600 ports and update the current list with the one given in the
603 Note that ip_local_port_range and ip_local_reserved_ports
604 settings are independent and both are considered by the kernel
605 when determining which ports are available for automatic port
608 You can reserve ports which are not in the current
609 ip_local_port_range, e.g.:
611 $ cat /proc/sys/net/ipv4/ip_local_port_range
613 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
616 although this is redundant. However such a setting is useful
617 if later the port range is changed to a value that will
618 include the reserved ports.
622 ip_nonlocal_bind - BOOLEAN
623 If set, allows processes to bind() to non-local IP addresses,
624 which can be quite useful - but may break some applications.
628 If set non-zero, enables support for dynamic addresses.
629 If set to a non-zero value larger than 1, a kernel log
630 message will be printed when dynamic address rewriting
634 icmp_echo_ignore_all - BOOLEAN
635 If set non-zero, then the kernel will ignore all ICMP ECHO
639 icmp_echo_ignore_broadcasts - BOOLEAN
640 If set non-zero, then the kernel will ignore all ICMP ECHO and
641 TIMESTAMP requests sent to it via broadcast/multicast.
644 icmp_ratelimit - INTEGER
645 Limit the maximal rates for sending ICMP packets whose type matches
646 icmp_ratemask (see below) to specific targets.
647 0 to disable any limiting,
648 otherwise the minimal space between responses in milliseconds.
651 icmp_ratemask - INTEGER
652 Mask made of ICMP types for which rates are being limited.
653 Significant bits: IHGFEDCBA9876543210
654 Default mask: 0000001100000011000 (6168)
656 Bit definitions (see include/linux/icmp.h):
658 3 Destination Unreachable *
663 C Parameter Problem *
668 H Address Mask Request
671 * These are rate limited by default (see default mask above)
673 icmp_ignore_bogus_error_responses - BOOLEAN
674 Some routers violate RFC1122 by sending bogus responses to broadcast
675 frames. Such violations are normally logged via a kernel warning.
676 If this is set to TRUE, the kernel will not give such warnings, which
677 will avoid log file clutter.
680 icmp_errors_use_inbound_ifaddr - BOOLEAN
682 If zero, icmp error messages are sent with the primary address of
683 the exiting interface.
685 If non-zero, the message will be sent with the primary address of
686 the interface that received the packet that caused the icmp error.
687 This is the behaviour network many administrators will expect from
688 a router. And it can make debugging complicated network layouts
691 Note that if no primary address exists for the interface selected,
692 then the primary address of the first non-loopback interface that
693 has one will be used regardless of this setting.
697 igmp_max_memberships - INTEGER
698 Change the maximum number of multicast groups we can subscribe to.
701 conf/interface/* changes special settings per interface (where "interface" is
702 the name of your network interface)
703 conf/all/* is special, changes the settings for all interfaces
706 log_martians - BOOLEAN
707 Log packets with impossible addresses to kernel log.
708 log_martians for the interface will be enabled if at least one of
709 conf/{all,interface}/log_martians is set to TRUE,
710 it will be disabled otherwise
712 accept_redirects - BOOLEAN
713 Accept ICMP redirect messages.
714 accept_redirects for the interface will be enabled if:
715 - both conf/{all,interface}/accept_redirects are TRUE in the case
716 forwarding for the interface is enabled
718 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
719 case forwarding for the interface is disabled
720 accept_redirects for the interface will be disabled otherwise
725 Enable IP forwarding on this interface.
727 mc_forwarding - BOOLEAN
728 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
729 and a multicast routing daemon is required.
730 conf/all/mc_forwarding must also be set to TRUE to enable multicast
731 routing for the interface
734 Integer value used to differentiate the devices by the medium they
735 are attached to. Two devices can have different id values when
736 the broadcast packets are received only on one of them.
737 The default value 0 means that the device is the only interface
738 to its medium, value of -1 means that medium is not known.
740 Currently, it is used to change the proxy_arp behavior:
741 the proxy_arp feature is enabled for packets forwarded between
742 two devices attached to different media.
746 proxy_arp for the interface will be enabled if at least one of
747 conf/{all,interface}/proxy_arp is set to TRUE,
748 it will be disabled otherwise
750 proxy_arp_pvlan - BOOLEAN
751 Private VLAN proxy arp.
752 Basically allow proxy arp replies back to the same interface
753 (from which the ARP request/solicitation was received).
755 This is done to support (ethernet) switch features, like RFC
756 3069, where the individual ports are NOT allowed to
757 communicate with each other, but they are allowed to talk to
758 the upstream router. As described in RFC 3069, it is possible
759 to allow these hosts to communicate through the upstream
760 router by proxy_arp'ing. Don't need to be used together with
763 This technology is known by different names:
764 In RFC 3069 it is called VLAN Aggregation.
765 Cisco and Allied Telesyn call it Private VLAN.
766 Hewlett-Packard call it Source-Port filtering or port-isolation.
767 Ericsson call it MAC-Forced Forwarding (RFC Draft).
769 shared_media - BOOLEAN
770 Send(router) or accept(host) RFC1620 shared media redirects.
771 Overrides ip_secure_redirects.
772 shared_media for the interface will be enabled if at least one of
773 conf/{all,interface}/shared_media is set to TRUE,
774 it will be disabled otherwise
777 secure_redirects - BOOLEAN
778 Accept ICMP redirect messages only for gateways,
779 listed in default gateway list.
780 secure_redirects for the interface will be enabled if at least one of
781 conf/{all,interface}/secure_redirects is set to TRUE,
782 it will be disabled otherwise
785 send_redirects - BOOLEAN
786 Send redirects, if router.
787 send_redirects for the interface will be enabled if at least one of
788 conf/{all,interface}/send_redirects is set to TRUE,
789 it will be disabled otherwise
792 bootp_relay - BOOLEAN
793 Accept packets with source address 0.b.c.d destined
794 not to this host as local ones. It is supposed, that
795 BOOTP relay daemon will catch and forward such packets.
796 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
801 accept_source_route - BOOLEAN
802 Accept packets with SRR option.
803 conf/all/accept_source_route must also be set to TRUE to accept packets
804 with SRR option on the interface
805 default TRUE (router)
808 accept_local - BOOLEAN
809 Accept packets with local source addresses. In combination with
810 suitable routing, this can be used to direct packets between two
811 local interfaces over the wire and have them accepted properly.
815 0 - No source validation.
816 1 - Strict mode as defined in RFC3704 Strict Reverse Path
817 Each incoming packet is tested against the FIB and if the interface
818 is not the best reverse path the packet check will fail.
819 By default failed packets are discarded.
820 2 - Loose mode as defined in RFC3704 Loose Reverse Path
821 Each incoming packet's source address is also tested against the FIB
822 and if the source address is not reachable via any interface
823 the packet check will fail.
825 Current recommended practice in RFC3704 is to enable strict mode
826 to prevent IP spoofing from DDos attacks. If using asymmetric routing
827 or other complicated routing, then loose mode is recommended.
829 The max value from conf/{all,interface}/rp_filter is used
830 when doing source validation on the {interface}.
832 Default value is 0. Note that some distributions enable it
836 1 - Allows you to have multiple network interfaces on the same
837 subnet, and have the ARPs for each interface be answered
838 based on whether or not the kernel would route a packet from
839 the ARP'd IP out that interface (therefore you must use source
840 based routing for this to work). In other words it allows control
841 of which cards (usually 1) will respond to an arp request.
843 0 - (default) The kernel can respond to arp requests with addresses
844 from other interfaces. This may seem wrong but it usually makes
845 sense, because it increases the chance of successful communication.
846 IP addresses are owned by the complete host on Linux, not by
847 particular interfaces. Only for more complex setups like load-
848 balancing, does this behaviour cause problems.
850 arp_filter for the interface will be enabled if at least one of
851 conf/{all,interface}/arp_filter is set to TRUE,
852 it will be disabled otherwise
854 arp_announce - INTEGER
855 Define different restriction levels for announcing the local
856 source IP address from IP packets in ARP requests sent on
858 0 - (default) Use any local address, configured on any interface
859 1 - Try to avoid local addresses that are not in the target's
860 subnet for this interface. This mode is useful when target
861 hosts reachable via this interface require the source IP
862 address in ARP requests to be part of their logical network
863 configured on the receiving interface. When we generate the
864 request we will check all our subnets that include the
865 target IP and will preserve the source address if it is from
866 such subnet. If there is no such subnet we select source
867 address according to the rules for level 2.
868 2 - Always use the best local address for this target.
869 In this mode we ignore the source address in the IP packet
870 and try to select local address that we prefer for talks with
871 the target host. Such local address is selected by looking
872 for primary IP addresses on all our subnets on the outgoing
873 interface that include the target IP address. If no suitable
874 local address is found we select the first local address
875 we have on the outgoing interface or on all other interfaces,
876 with the hope we will receive reply for our request and
877 even sometimes no matter the source IP address we announce.
879 The max value from conf/{all,interface}/arp_announce is used.
881 Increasing the restriction level gives more chance for
882 receiving answer from the resolved target while decreasing
883 the level announces more valid sender's information.
886 Define different modes for sending replies in response to
887 received ARP requests that resolve local target IP addresses:
888 0 - (default): reply for any local target IP address, configured
890 1 - reply only if the target IP address is local address
891 configured on the incoming interface
892 2 - reply only if the target IP address is local address
893 configured on the incoming interface and both with the
894 sender's IP address are part from same subnet on this interface
895 3 - do not reply for local addresses configured with scope host,
896 only resolutions for global and link addresses are replied
898 8 - do not reply for all local addresses
900 The max value from conf/{all,interface}/arp_ignore is used
901 when ARP request is received on the {interface}
904 Define mode for notification of address and device changes.
905 0 - (default): do nothing
906 1 - Generate gratuitous arp replies when device is brought up
907 or hardware address changes.
910 Define behavior for gratuitous ARP frames who's IP is not
911 already present in the ARP table:
912 0 - don't create new entries in the ARP table
913 1 - create new entries in the ARP table
915 Both replies and requests type gratuitous arp will trigger the
916 ARP table to be updated, if this setting is on.
918 If the ARP table already contains the IP address of the
919 gratuitous arp frame, the arp table will be updated regardless
920 if this setting is on or off.
923 app_solicit - INTEGER
924 The maximum number of probes to send to the user space ARP daemon
925 via netlink before dropping back to multicast probes (see
926 mcast_solicit). Defaults to 0.
928 disable_policy - BOOLEAN
929 Disable IPSEC policy (SPD) for this interface
931 disable_xfrm - BOOLEAN
932 Disable IPSEC encryption on this interface, whatever the policy
937 Allows you to write a number, which can be used as required.
947 delon.nicolas@wanadoo.fr
952 /proc/sys/net/ipv6/* Variables:
954 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
955 apply to IPv6 [XXX?].
958 Default value for IPV6_V6ONLY socket option,
959 which restricts use of the IPv6 socket to IPv6 communication
961 TRUE: disable IPv4-mapped address feature
962 FALSE: enable IPv4-mapped address feature
964 Default: FALSE (as specified in RFC2553bis)
968 ip6frag_high_thresh - INTEGER
969 Maximum memory used to reassemble IPv6 fragments. When
970 ip6frag_high_thresh bytes of memory is allocated for this purpose,
971 the fragment handler will toss packets until ip6frag_low_thresh
974 ip6frag_low_thresh - INTEGER
975 See ip6frag_high_thresh
977 ip6frag_time - INTEGER
978 Time in seconds to keep an IPv6 fragment in memory.
980 ip6frag_secret_interval - INTEGER
981 Regeneration interval (in seconds) of the hash secret (or lifetime
982 for the hash secret) for IPv6 fragments.
986 Change the interface-specific default settings.
990 Change all the interface-specific settings.
992 [XXX: Other special features than forwarding?]
994 conf/all/forwarding - BOOLEAN
995 Enable global IPv6 forwarding between all interfaces.
997 IPv4 and IPv6 work differently here; e.g. netfilter must be used
998 to control which interfaces may forward packets and which not.
1000 This also sets all interfaces' Host/Router setting
1001 'forwarding' to the specified value. See below for details.
1003 This referred to as global forwarding.
1009 Change special settings per interface.
1011 The functional behaviour for certain settings is different
1012 depending on whether local forwarding is enabled or not.
1015 Accept Router Advertisements; autoconfigure using them.
1017 Functional default: enabled if local forwarding is disabled.
1018 disabled if local forwarding is enabled.
1020 accept_ra_defrtr - BOOLEAN
1021 Learn default router in Router Advertisement.
1023 Functional default: enabled if accept_ra is enabled.
1024 disabled if accept_ra is disabled.
1026 accept_ra_pinfo - BOOLEAN
1027 Learn Prefix Information in Router Advertisement.
1029 Functional default: enabled if accept_ra is enabled.
1030 disabled if accept_ra is disabled.
1032 accept_ra_rt_info_max_plen - INTEGER
1033 Maximum prefix length of Route Information in RA.
1035 Route Information w/ prefix larger than or equal to this
1036 variable shall be ignored.
1038 Functional default: 0 if accept_ra_rtr_pref is enabled.
1039 -1 if accept_ra_rtr_pref is disabled.
1041 accept_ra_rtr_pref - BOOLEAN
1042 Accept Router Preference in RA.
1044 Functional default: enabled if accept_ra is enabled.
1045 disabled if accept_ra is disabled.
1047 accept_redirects - BOOLEAN
1050 Functional default: enabled if local forwarding is disabled.
1051 disabled if local forwarding is enabled.
1053 accept_source_route - INTEGER
1054 Accept source routing (routing extension header).
1056 >= 0: Accept only routing header type 2.
1057 < 0: Do not accept routing header.
1062 Autoconfigure addresses using Prefix Information in Router
1065 Functional default: enabled if accept_ra_pinfo is enabled.
1066 disabled if accept_ra_pinfo is disabled.
1068 dad_transmits - INTEGER
1069 The amount of Duplicate Address Detection probes to send.
1072 forwarding - BOOLEAN
1073 Configure interface-specific Host/Router behaviour.
1075 Note: It is recommended to have the same setting on all
1076 interfaces; mixed router/host scenarios are rather uncommon.
1080 By default, Host behaviour is assumed. This means:
1082 1. IsRouter flag is not set in Neighbour Advertisements.
1083 2. Router Solicitations are being sent when necessary.
1084 3. If accept_ra is TRUE (default), accept Router
1085 Advertisements (and do autoconfiguration).
1086 4. If accept_redirects is TRUE (default), accept Redirects.
1090 If local forwarding is enabled, Router behaviour is assumed.
1091 This means exactly the reverse from the above:
1093 1. IsRouter flag is set in Neighbour Advertisements.
1094 2. Router Solicitations are not sent.
1095 3. Router Advertisements are ignored.
1096 4. Redirects are ignored.
1098 Default: FALSE if global forwarding is disabled (default),
1102 Default Hop Limit to set.
1106 Default Maximum Transfer Unit
1107 Default: 1280 (IPv6 required minimum)
1109 router_probe_interval - INTEGER
1110 Minimum interval (in seconds) between Router Probing described
1115 router_solicitation_delay - INTEGER
1116 Number of seconds to wait after interface is brought up
1117 before sending Router Solicitations.
1120 router_solicitation_interval - INTEGER
1121 Number of seconds to wait between Router Solicitations.
1124 router_solicitations - INTEGER
1125 Number of Router Solicitations to send until assuming no
1126 routers are present.
1129 use_tempaddr - INTEGER
1130 Preference for Privacy Extensions (RFC3041).
1131 <= 0 : disable Privacy Extensions
1132 == 1 : enable Privacy Extensions, but prefer public
1133 addresses over temporary addresses.
1134 > 1 : enable Privacy Extensions and prefer temporary
1135 addresses over public addresses.
1136 Default: 0 (for most devices)
1137 -1 (for point-to-point devices and loopback devices)
1139 temp_valid_lft - INTEGER
1140 valid lifetime (in seconds) for temporary addresses.
1141 Default: 604800 (7 days)
1143 temp_prefered_lft - INTEGER
1144 Preferred lifetime (in seconds) for temporary addresses.
1145 Default: 86400 (1 day)
1147 max_desync_factor - INTEGER
1148 Maximum value for DESYNC_FACTOR, which is a random value
1149 that ensures that clients don't synchronize with each
1150 other and generate new addresses at exactly the same time.
1151 value is in seconds.
1154 regen_max_retry - INTEGER
1155 Number of attempts before give up attempting to generate
1156 valid temporary addresses.
1159 max_addresses - INTEGER
1160 Maximum number of autoconfigured addresses per interface. Setting
1161 to zero disables the limitation. It is not recommended to set this
1162 value too large (or to zero) because it would be an easy way to
1163 crash the kernel by allowing too many addresses to be created.
1166 disable_ipv6 - BOOLEAN
1167 Disable IPv6 operation. If accept_dad is set to 2, this value
1168 will be dynamically set to TRUE if DAD fails for the link-local
1170 Default: FALSE (enable IPv6 operation)
1172 When this value is changed from 1 to 0 (IPv6 is being enabled),
1173 it will dynamically create a link-local address on the given
1174 interface and start Duplicate Address Detection, if necessary.
1176 When this value is changed from 0 to 1 (IPv6 is being disabled),
1177 it will dynamically delete all address on the given interface.
1179 accept_dad - INTEGER
1180 Whether to accept DAD (Duplicate Address Detection).
1182 1: Enable DAD (default)
1183 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1184 link-local address has been found.
1186 force_tllao - BOOLEAN
1187 Enable sending the target link-layer address option even when
1188 responding to a unicast neighbor solicitation.
1191 Quoting from RFC 2461, section 4.4, Target link-layer address:
1193 "The option MUST be included for multicast solicitations in order to
1194 avoid infinite Neighbor Solicitation "recursion" when the peer node
1195 does not have a cache entry to return a Neighbor Advertisements
1196 message. When responding to unicast solicitations, the option can be
1197 omitted since the sender of the solicitation has the correct link-
1198 layer address; otherwise it would not have be able to send the unicast
1199 solicitation in the first place. However, including the link-layer
1200 address in this case adds little overhead and eliminates a potential
1201 race condition where the sender deletes the cached link-layer address
1202 prior to receiving a response to a previous solicitation."
1206 Limit the maximal rates for sending ICMPv6 packets.
1207 0 to disable any limiting,
1208 otherwise the minimal space between responses in milliseconds.
1213 Pekka Savola <pekkas@netcore.fi>
1214 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1217 /proc/sys/net/bridge/* Variables:
1219 bridge-nf-call-arptables - BOOLEAN
1220 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1224 bridge-nf-call-iptables - BOOLEAN
1225 1 : pass bridged IPv4 traffic to iptables' chains.
1229 bridge-nf-call-ip6tables - BOOLEAN
1230 1 : pass bridged IPv6 traffic to ip6tables' chains.
1234 bridge-nf-filter-vlan-tagged - BOOLEAN
1235 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1239 bridge-nf-filter-pppoe-tagged - BOOLEAN
1240 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1245 proc/sys/net/sctp/* Variables:
1247 addip_enable - BOOLEAN
1248 Enable or disable extension of Dynamic Address Reconfiguration
1249 (ADD-IP) functionality specified in RFC5061. This extension provides
1250 the ability to dynamically add and remove new addresses for the SCTP
1253 1: Enable extension.
1255 0: Disable extension.
1259 addip_noauth_enable - BOOLEAN
1260 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1261 authentication to protect the operations of adding or removing new
1262 addresses. This requirement is mandated so that unauthorized hosts
1263 would not be able to hijack associations. However, older
1264 implementations may not have implemented this requirement while
1265 allowing the ADD-IP extension. For reasons of interoperability,
1266 we provide this variable to control the enforcement of the
1267 authentication requirement.
1269 1: Allow ADD-IP extension to be used without authentication. This
1270 should only be set in a closed environment for interoperability
1271 with older implementations.
1273 0: Enforce the authentication requirement
1277 auth_enable - BOOLEAN
1278 Enable or disable Authenticated Chunks extension. This extension
1279 provides the ability to send and receive authenticated chunks and is
1280 required for secure operation of Dynamic Address Reconfiguration
1283 1: Enable this extension.
1284 0: Disable this extension.
1288 prsctp_enable - BOOLEAN
1289 Enable or disable the Partial Reliability extension (RFC3758) which
1290 is used to notify peers that a given DATA should no longer be expected.
1298 The limit of the number of new packets that can be initially sent. It
1299 controls how bursty the generated traffic can be.
1303 association_max_retrans - INTEGER
1304 Set the maximum number for retransmissions that an association can
1305 attempt deciding that the remote end is unreachable. If this value
1306 is exceeded, the association is terminated.
1310 max_init_retransmits - INTEGER
1311 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1312 that an association will attempt before declaring the destination
1313 unreachable and terminating.
1317 path_max_retrans - INTEGER
1318 The maximum number of retransmissions that will be attempted on a given
1319 path. Once this threshold is exceeded, the path is considered
1320 unreachable, and new traffic will use a different path when the
1321 association is multihomed.
1325 rto_initial - INTEGER
1326 The initial round trip timeout value in milliseconds that will be used
1327 in calculating round trip times. This is the initial time interval
1328 for retransmissions.
1333 The maximum value (in milliseconds) of the round trip timeout. This
1334 is the largest time interval that can elapse between retransmissions.
1339 The minimum value (in milliseconds) of the round trip timeout. This
1340 is the smallest time interval the can elapse between retransmissions.
1344 hb_interval - INTEGER
1345 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1346 are sent at the specified interval on idle paths to probe the state of
1347 a given path between 2 associations.
1351 sack_timeout - INTEGER
1352 The amount of time (in milliseconds) that the implementation will wait
1357 valid_cookie_life - INTEGER
1358 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1359 is used during association establishment.
1363 cookie_preserve_enable - BOOLEAN
1364 Enable or disable the ability to extend the lifetime of the SCTP cookie
1365 that is used during the establishment phase of SCTP association
1367 1: Enable cookie lifetime extension.
1372 rcvbuf_policy - INTEGER
1373 Determines if the receive buffer is attributed to the socket or to
1374 association. SCTP supports the capability to create multiple
1375 associations on a single socket. When using this capability, it is
1376 possible that a single stalled association that's buffering a lot
1377 of data may block other associations from delivering their data by
1378 consuming all of the receive buffer space. To work around this,
1379 the rcvbuf_policy could be set to attribute the receiver buffer space
1380 to each association instead of the socket. This prevents the described
1383 1: rcvbuf space is per association
1384 0: recbuf space is per socket
1388 sndbuf_policy - INTEGER
1389 Similar to rcvbuf_policy above, this applies to send buffer space.
1391 1: Send buffer is tracked per association
1392 0: Send buffer is tracked per socket.
1396 sctp_mem - vector of 3 INTEGERs: min, pressure, max
1397 Number of pages allowed for queueing by all SCTP sockets.
1399 min: Below this number of pages SCTP is not bothered about its
1400 memory appetite. When amount of memory allocated by SCTP exceeds
1401 this number, SCTP starts to moderate memory usage.
1403 pressure: This value was introduced to follow format of tcp_mem.
1405 max: Number of pages allowed for queueing by all SCTP sockets.
1407 Default is calculated at boot time from amount of available memory.
1409 sctp_rmem - vector of 3 INTEGERs: min, default, max
1410 See tcp_rmem for a description.
1412 sctp_wmem - vector of 3 INTEGERs: min, default, max
1413 See tcp_wmem for a description.
1415 addr_scope_policy - INTEGER
1416 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1418 0 - Disable IPv4 address scoping
1419 1 - Enable IPv4 address scoping
1420 2 - Follow draft but allow IPv4 private addresses
1421 3 - Follow draft but allow IPv4 link local addresses
1426 /proc/sys/net/core/*
1427 dev_weight - INTEGER
1428 The maximum number of packets that kernel can handle on a NAPI
1429 interrupt, it's a Per-CPU variable.
1433 /proc/sys/net/unix/*
1434 max_dgram_qlen - INTEGER
1435 The maximum length of dgram socket receive queue
1442 /proc/sys/net/irda/*
1443 fast_poll_increase FIXME
1444 warn_noreply_time FIXME
1445 discovery_slots FIXME
1448 discovery_timeout FIXME
1449 lap_keepalive_time FIXME
1450 max_noreply_time FIXME
1451 max_tx_data_size FIXME
1453 min_tx_turn_time FIXME