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.
22 ipfrag_high_thresh - INTEGER
23 Maximum memory used to reassemble IP fragments. When
24 ipfrag_high_thresh bytes of memory is allocated for this purpose,
25 the fragment handler will toss packets until ipfrag_low_thresh
28 ipfrag_low_thresh - INTEGER
29 See ipfrag_high_thresh
32 Time in seconds to keep an IP fragment in memory.
34 ipfrag_secret_interval - INTEGER
35 Regeneration interval (in seconds) of the hash secret (or lifetime
36 for the hash secret) for IP fragments.
41 inet_peer_threshold - INTEGER
42 The approximate size of the storage. Starting from this threshold
43 entries will be thrown aggressively. This threshold also determines
44 entries' time-to-live and time intervals between garbage collection
45 passes. More entries, less time-to-live, less GC interval.
47 inet_peer_minttl - INTEGER
48 Minimum time-to-live of entries. Should be enough to cover fragment
49 time-to-live on the reassembling side. This minimum time-to-live is
50 guaranteed if the pool size is less than inet_peer_threshold.
51 Measured in jiffies(1).
53 inet_peer_maxttl - INTEGER
54 Maximum time-to-live of entries. Unused entries will expire after
55 this period of time if there is no memory pressure on the pool (i.e.
56 when the number of entries in the pool is very small).
57 Measured in jiffies(1).
59 inet_peer_gc_mintime - INTEGER
60 Minimum interval between garbage collection passes. This interval is
61 in effect under high memory pressure on the pool.
62 Measured in jiffies(1).
64 inet_peer_gc_maxtime - INTEGER
65 Minimum interval between garbage collection passes. This interval is
66 in effect under low (or absent) memory pressure on the pool.
67 Measured in jiffies(1).
71 tcp_syn_retries - INTEGER
72 Number of times initial SYNs for an active TCP connection attempt
73 will be retransmitted. Should not be higher than 255. Default value
74 is 5, which corresponds to ~180seconds.
76 tcp_synack_retries - INTEGER
77 Number of times SYNACKs for a passive TCP connection attempt will
78 be retransmitted. Should not be higher than 255. Default value
79 is 5, which corresponds to ~180seconds.
81 tcp_keepalive_time - INTEGER
82 How often TCP sends out keepalive messages when keepalive is enabled.
85 tcp_keepalive_probes - INTEGER
86 How many keepalive probes TCP sends out, until it decides that the
87 connection is broken. Default value: 9.
89 tcp_keepalive_intvl - INTEGER
90 How frequently the probes are send out. Multiplied by
91 tcp_keepalive_probes it is time to kill not responding connection,
92 after probes started. Default value: 75sec i.e. connection
93 will be aborted after ~11 minutes of retries.
95 tcp_retries1 - INTEGER
96 How many times to retry before deciding that something is wrong
97 and it is necessary to report this suspicion to network layer.
98 Minimal RFC value is 3, it is default, which corresponds
99 to ~3sec-8min depending on RTO.
101 tcp_retries2 - INTEGER
102 How may times to retry before killing alive TCP connection.
103 RFC1122 says that the limit should be longer than 100 sec.
104 It is too small number. Default value 15 corresponds to ~13-30min
107 tcp_orphan_retries - INTEGER
108 How may times to retry before killing TCP connection, closed
109 by our side. Default value 7 corresponds to ~50sec-16min
110 depending on RTO. If you machine is loaded WEB server,
111 you should think about lowering this value, such sockets
112 may consume significant resources. Cf. tcp_max_orphans.
114 tcp_fin_timeout - INTEGER
115 Time to hold socket in state FIN-WAIT-2, if it was closed
116 by our side. Peer can be broken and never close its side,
117 or even died unexpectedly. Default value is 60sec.
118 Usual value used in 2.2 was 180 seconds, you may restore
119 it, but remember that if your machine is even underloaded WEB server,
120 you risk to overflow memory with kilotons of dead sockets,
121 FIN-WAIT-2 sockets are less dangerous than FIN-WAIT-1,
122 because they eat maximum 1.5K of memory, but they tend
123 to live longer. Cf. tcp_max_orphans.
125 tcp_max_tw_buckets - INTEGER
126 Maximal number of timewait sockets held by system simultaneously.
127 If this number is exceeded time-wait socket is immediately destroyed
128 and warning is printed. This limit exists only to prevent
129 simple DoS attacks, you _must_ not lower the limit artificially,
130 but rather increase it (probably, after increasing installed memory),
131 if network conditions require more than default value.
133 tcp_tw_recycle - BOOLEAN
134 Enable fast recycling TIME-WAIT sockets. Default value is 0.
135 It should not be changed without advice/request of technical
138 tcp_tw_reuse - BOOLEAN
139 Allow to reuse TIME-WAIT sockets for new connections when it is
140 safe from protocol viewpoint. Default value is 0.
141 It should not be changed without advice/request of technical
144 tcp_max_orphans - INTEGER
145 Maximal number of TCP sockets not attached to any user file handle,
146 held by system. If this number is exceeded orphaned connections are
147 reset immediately and warning is printed. This limit exists
148 only to prevent simple DoS attacks, you _must_ not rely on this
149 or lower the limit artificially, but rather increase it
150 (probably, after increasing installed memory),
151 if network conditions require more than default value,
152 and tune network services to linger and kill such states
153 more aggressively. Let me to remind again: each orphan eats
154 up to ~64K of unswappable memory.
156 tcp_abort_on_overflow - BOOLEAN
157 If listening service is too slow to accept new connections,
158 reset them. Default state is FALSE. It means that if overflow
159 occurred due to a burst, connection will recover. Enable this
160 option _only_ if you are really sure that listening daemon
161 cannot be tuned to accept connections faster. Enabling this
162 option can harm clients of your server.
164 tcp_syncookies - BOOLEAN
165 Only valid when the kernel was compiled with CONFIG_SYNCOOKIES
166 Send out syncookies when the syn backlog queue of a socket
167 overflows. This is to prevent against the common 'syn flood attack'
170 Note, that syncookies is fallback facility.
171 It MUST NOT be used to help highly loaded servers to stand
172 against legal connection rate. If you see synflood warnings
173 in your logs, but investigation shows that they occur
174 because of overload with legal connections, you should tune
175 another parameters until this warning disappear.
176 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
178 syncookies seriously violate TCP protocol, do not allow
179 to use TCP extensions, can result in serious degradation
180 of some services (f.e. SMTP relaying), visible not by you,
181 but your clients and relays, contacting you. While you see
182 synflood warnings in logs not being really flooded, your server
183 is seriously misconfigured.
186 Use the Host requirements interpretation of the TCP urg pointer field.
187 Most hosts use the older BSD interpretation, so if you turn this on
188 Linux might not communicate correctly with them.
191 tcp_max_syn_backlog - INTEGER
192 Maximal number of remembered connection requests, which are
193 still did not receive an acknowledgment from connecting client.
194 Default value is 1024 for systems with more than 128Mb of memory,
195 and 128 for low memory machines. If server suffers of overload,
196 try to increase this number.
198 tcp_window_scaling - BOOLEAN
199 Enable window scaling as defined in RFC1323.
201 tcp_timestamps - BOOLEAN
202 Enable timestamps as defined in RFC1323.
205 Enable select acknowledgments (SACKS).
208 Enable FACK congestion avoidance and fast retransmission.
209 The value is not used, if tcp_sack is not enabled.
212 Allows TCP to send "duplicate" SACKs.
215 Enable Explicit Congestion Notification in TCP.
217 tcp_reordering - INTEGER
218 Maximal reordering of packets in a TCP stream.
221 tcp_retrans_collapse - BOOLEAN
222 Bug-to-bug compatibility with some broken printers.
223 On retransmit try to send bigger packets to work around bugs in
226 tcp_wmem - vector of 3 INTEGERs: min, default, max
227 min: Amount of memory reserved for send buffers for TCP socket.
228 Each TCP socket has rights to use it due to fact of its birth.
231 default: Amount of memory allowed for send buffers for TCP socket
232 by default. This value overrides net.core.wmem_default used
233 by other protocols, it is usually lower than net.core.wmem_default.
236 max: Maximal amount of memory allowed for automatically selected
237 send buffers for TCP socket. This value does not override
238 net.core.wmem_max, "static" selection via SO_SNDBUF does not use this.
241 tcp_rmem - vector of 3 INTEGERs: min, default, max
242 min: Minimal size of receive buffer used by TCP sockets.
243 It is guaranteed to each TCP socket, even under moderate memory
247 default: default size of receive buffer used by TCP sockets.
248 This value overrides net.core.rmem_default used by other protocols.
249 Default: 87380 bytes. This value results in window of 65535 with
250 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
251 less for default tcp_app_win. See below about these variables.
253 max: maximal size of receive buffer allowed for automatically
254 selected receiver buffers for TCP socket. This value does not override
255 net.core.rmem_max, "static" selection via SO_RCVBUF does not use this.
256 Default: 87380*2 bytes.
258 tcp_mem - vector of 3 INTEGERs: min, pressure, max
259 low: below this number of pages TCP is not bothered about its
262 pressure: when amount of memory allocated by TCP exceeds this number
263 of pages, TCP moderates its memory consumption and enters memory
264 pressure mode, which is exited when memory consumption falls
267 high: number of pages allowed for queueing by all TCP sockets.
269 Defaults are calculated at boot time from amount of available
272 tcp_app_win - INTEGER
273 Reserve max(window/2^tcp_app_win, mss) of window for application
274 buffer. Value 0 is special, it means that nothing is reserved.
277 tcp_adv_win_scale - INTEGER
278 Count buffering overhead as bytes/2^tcp_adv_win_scale
279 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
283 tcp_rfc1337 - BOOLEAN
284 If set, the TCP stack behaves conforming to RFC1337. If unset,
285 we are not conforming to RFC, but prevent TCP TIME_WAIT
289 tcp_low_latency - BOOLEAN
290 If set, the TCP stack makes decisions that prefer lower
291 latency as opposed to higher throughput. By default, this
292 option is not set meaning that higher throughput is preferred.
293 An example of an application where this default should be
294 changed would be a Beowulf compute cluster.
297 tcp_westwood - BOOLEAN
298 Enable TCP Westwood+ congestion control algorithm.
299 TCP Westwood+ is a sender-side only modification of the TCP Reno
300 protocol stack that optimizes the performance of TCP congestion
301 control. It is based on end-to-end bandwidth estimation to set
302 congestion window and slow start threshold after a congestion
303 episode. Using this estimation, TCP Westwood+ adaptively sets a
304 slow start threshold and a congestion window which takes into
305 account the bandwidth used at the time congestion is experienced.
306 TCP Westwood+ significantly increases fairness wrt TCP Reno in
307 wired networks and throughput over wireless links.
310 ip_local_port_range - 2 INTEGERS
311 Defines the local port range that is used by TCP and UDP to
312 choose the local port. The first number is the first, the
313 second the last local port number. Default value depends on
314 amount of memory available on the system:
316 < 128Mb 1024-4999 or even less.
317 This number defines number of active connections, which this
318 system can issue simultaneously to systems not supporting
319 TCP extensions (timestamps). With tcp_tw_recycle enabled
320 (i.e. by default) range 1024-4999 is enough to issue up to
321 2000 connections per second to systems supporting timestamps.
323 ip_nonlocal_bind - BOOLEAN
324 If set, allows processes to bind() to non-local IP addresses,
325 which can be quite useful - but may break some applications.
329 If set non-zero, enables support for dynamic addresses.
330 If set to a non-zero value larger than 1, a kernel log
331 message will be printed when dynamic address rewriting
335 icmp_echo_ignore_all - BOOLEAN
336 icmp_echo_ignore_broadcasts - BOOLEAN
337 If either is set to true, then the kernel will ignore either all
338 ICMP ECHO requests sent to it or just those to broadcast/multicast
339 addresses, respectively.
341 icmp_ratelimit - INTEGER
342 Limit the maximal rates for sending ICMP packets whose type matches
343 icmp_ratemask (see below) to specific targets.
344 0 to disable any limiting, otherwise the maximal rate in jiffies(1)
347 icmp_ratemask - INTEGER
348 Mask made of ICMP types for which rates are being limited.
349 Significant bits: IHGFEDCBA9876543210
350 Default mask: 0000001100000011000 (6168)
352 Bit definitions (see include/linux/icmp.h):
354 3 Destination Unreachable *
359 C Parameter Problem *
364 H Address Mask Request
367 * These are rate limited by default (see default mask above)
369 icmp_ignore_bogus_error_responses - BOOLEAN
370 Some routers violate RFC1122 by sending bogus responses to broadcast
371 frames. Such violations are normally logged via a kernel warning.
372 If this is set to TRUE, the kernel will not give such warnings, which
373 will avoid log file clutter.
376 igmp_max_memberships - INTEGER
377 Change the maximum number of multicast groups we can subscribe to.
380 conf/interface/* changes special settings per interface (where "interface" is
381 the name of your network interface)
382 conf/all/* is special, changes the settings for all interfaces
385 log_martians - BOOLEAN
386 Log packets with impossible addresses to kernel log.
387 log_martians for the interface will be enabled if at least one of
388 conf/{all,interface}/log_martians is set to TRUE,
389 it will be disabled otherwise
391 accept_redirects - BOOLEAN
392 Accept ICMP redirect messages.
393 accept_redirects for the interface will be enabled if:
394 - both conf/{all,interface}/accept_redirects are TRUE in the case forwarding
395 for the interface is enabled
397 - at least one of conf/{all,interface}/accept_redirects is TRUE in the case
398 forwarding for the interface is disabled
399 accept_redirects for the interface will be disabled otherwise
404 Enable IP forwarding on this interface.
406 mc_forwarding - BOOLEAN
407 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
408 and a multicast routing daemon is required.
409 conf/all/mc_forwarding must also be set to TRUE to enable multicast routing
413 Integer value used to differentiate the devices by the medium they
414 are attached to. Two devices can have different id values when
415 the broadcast packets are received only on one of them.
416 The default value 0 means that the device is the only interface
417 to its medium, value of -1 means that medium is not known.
419 Currently, it is used to change the proxy_arp behavior:
420 the proxy_arp feature is enabled for packets forwarded between
421 two devices attached to different media.
425 proxy_arp for the interface will be enabled if at least one of
426 conf/{all,interface}/proxy_arp is set to TRUE,
427 it will be disabled otherwise
429 shared_media - BOOLEAN
430 Send(router) or accept(host) RFC1620 shared media redirects.
431 Overrides ip_secure_redirects.
432 shared_media for the interface will be enabled if at least one of
433 conf/{all,interface}/shared_media is set to TRUE,
434 it will be disabled otherwise
437 secure_redirects - BOOLEAN
438 Accept ICMP redirect messages only for gateways,
439 listed in default gateway list.
440 secure_redirects for the interface will be enabled if at least one of
441 conf/{all,interface}/secure_redirects is set to TRUE,
442 it will be disabled otherwise
445 send_redirects - BOOLEAN
446 Send redirects, if router.
447 send_redirects for the interface will be enabled if at least one of
448 conf/{all,interface}/send_redirects is set to TRUE,
449 it will be disabled otherwise
452 bootp_relay - BOOLEAN
453 Accept packets with source address 0.b.c.d destined
454 not to this host as local ones. It is supposed, that
455 BOOTP relay daemon will catch and forward such packets.
456 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
461 accept_source_route - BOOLEAN
462 Accept packets with SRR option.
463 conf/all/accept_source_route must also be set to TRUE to accept packets
464 with SRR option on the interface
465 default TRUE (router)
469 1 - do source validation by reversed path, as specified in RFC1812
470 Recommended option for single homed hosts and stub network
471 routers. Could cause troubles for complicated (not loop free)
472 networks running a slow unreliable protocol (sort of RIP),
473 or using static routes.
475 0 - No source validation.
477 conf/all/rp_filter must also be set to TRUE to do source validation
480 Default value is 0. Note that some distributions enable it
484 1 - Allows you to have multiple network interfaces on the same
485 subnet, and have the ARPs for each interface be answered
486 based on whether or not the kernel would route a packet from
487 the ARP'd IP out that interface (therefore you must use source
488 based routing for this to work). In other words it allows control
489 of which cards (usually 1) will respond to an arp request.
491 0 - (default) The kernel can respond to arp requests with addresses
492 from other interfaces. This may seem wrong but it usually makes
493 sense, because it increases the chance of successful communication.
494 IP addresses are owned by the complete host on Linux, not by
495 particular interfaces. Only for more complex setups like load-
496 balancing, does this behaviour cause problems.
498 arp_filter for the interface will be enabled if at least one of
499 conf/{all,interface}/arp_filter is set to TRUE,
500 it will be disabled otherwise
502 arp_announce - INTEGER
503 Define different restriction levels for announcing the local
504 source IP address from IP packets in ARP requests sent on
506 0 - (default) Use any local address, configured on any interface
507 1 - Try to avoid local addresses that are not in the target's
508 subnet for this interface. This mode is useful when target
509 hosts reachable via this interface require the source IP
510 address in ARP requests to be part of their logical network
511 configured on the receiving interface. When we generate the
512 request we will check all our subnets that include the
513 target IP and will preserve the source address if it is from
514 such subnet. If there is no such subnet we select source
515 address according to the rules for level 2.
516 2 - Always use the best local address for this target.
517 In this mode we ignore the source address in the IP packet
518 and try to select local address that we prefer for talks with
519 the target host. Such local address is selected by looking
520 for primary IP addresses on all our subnets on the outgoing
521 interface that include the target IP address. If no suitable
522 local address is found we select the first local address
523 we have on the outgoing interface or on all other interfaces,
524 with the hope we will receive reply for our request and
525 even sometimes no matter the source IP address we announce.
527 The max value from conf/{all,interface}/arp_announce is used.
529 Increasing the restriction level gives more chance for
530 receiving answer from the resolved target while decreasing
531 the level announces more valid sender's information.
534 Define different modes for sending replies in response to
535 received ARP requests that resolve local target IP addresses:
536 0 - (default): reply for any local target IP address, configured
538 1 - reply only if the target IP address is local address
539 configured on the incoming interface
540 2 - reply only if the target IP address is local address
541 configured on the incoming interface and both with the
542 sender's IP address are part from same subnet on this interface
543 3 - do not reply for local addresses configured with scope host,
544 only resolutions for global and link addresses are replied
546 8 - do not reply for all local addresses
548 The max value from conf/{all,interface}/arp_ignore is used
549 when ARP request is received on the {interface}
552 Allows you to write a number, which can be used as required.
555 (1) Jiffie: internal timeunit for the kernel. On the i386 1/100s, on the
556 Alpha 1/1024s. See the HZ define in /usr/include/asm/param.h for the exact
557 value on your system.
566 delon.nicolas@wanadoo.fr
571 /proc/sys/net/ipv6/* Variables:
573 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
574 apply to IPv6 [XXX?].
577 Default value for IPV6_V6ONLY socket option,
578 which restricts use of the IPv6 socket to IPv6 communication
580 TRUE: disable IPv4-mapped address feature
581 FALSE: enable IPv4-mapped address feature
583 Default: FALSE (as specified in RFC2553bis)
587 ip6frag_high_thresh - INTEGER
588 Maximum memory used to reassemble IPv6 fragments. When
589 ip6frag_high_thresh bytes of memory is allocated for this purpose,
590 the fragment handler will toss packets until ip6frag_low_thresh
593 ip6frag_low_thresh - INTEGER
594 See ip6frag_high_thresh
596 ip6frag_time - INTEGER
597 Time in seconds to keep an IPv6 fragment in memory.
599 ip6frag_secret_interval - INTEGER
600 Regeneration interval (in seconds) of the hash secret (or lifetime
601 for the hash secret) for IPv6 fragments.
605 Change the interface-specific default settings.
609 Change all the interface-specific settings.
611 [XXX: Other special features than forwarding?]
613 conf/all/forwarding - BOOLEAN
614 Enable global IPv6 forwarding between all interfaces.
616 IPv4 and IPv6 work differently here; e.g. netfilter must be used
617 to control which interfaces may forward packets and which not.
619 This also sets all interfaces' Host/Router setting
620 'forwarding' to the specified value. See below for details.
622 This referred to as global forwarding.
625 Change special settings per interface.
627 The functional behaviour for certain settings is different
628 depending on whether local forwarding is enabled or not.
631 Accept Router Advertisements; autoconfigure using them.
633 Functional default: enabled if local forwarding is disabled.
634 disabled if local forwarding is enabled.
636 accept_redirects - BOOLEAN
639 Functional default: enabled if local forwarding is disabled.
640 disabled if local forwarding is enabled.
643 Configure link-local addresses using L2 hardware addresses.
647 dad_transmits - INTEGER
648 The amount of Duplicate Address Detection probes to send.
652 Configure interface-specific Host/Router behaviour.
654 Note: It is recommended to have the same setting on all
655 interfaces; mixed router/host scenarios are rather uncommon.
659 By default, Host behaviour is assumed. This means:
661 1. IsRouter flag is not set in Neighbour Advertisements.
662 2. Router Solicitations are being sent when necessary.
663 3. If accept_ra is TRUE (default), accept Router
664 Advertisements (and do autoconfiguration).
665 4. If accept_redirects is TRUE (default), accept Redirects.
669 If local forwarding is enabled, Router behaviour is assumed.
670 This means exactly the reverse from the above:
672 1. IsRouter flag is set in Neighbour Advertisements.
673 2. Router Solicitations are not sent.
674 3. Router Advertisements are ignored.
675 4. Redirects are ignored.
677 Default: FALSE if global forwarding is disabled (default),
681 Default Hop Limit to set.
685 Default Maximum Transfer Unit
686 Default: 1280 (IPv6 required minimum)
688 router_solicitation_delay - INTEGER
689 Number of seconds to wait after interface is brought up
690 before sending Router Solicitations.
693 router_solicitation_interval - INTEGER
694 Number of seconds to wait between Router Solicitations.
697 router_solicitations - INTEGER
698 Number of Router Solicitations to send until assuming no
702 use_tempaddr - INTEGER
703 Preference for Privacy Extensions (RFC3041).
704 <= 0 : disable Privacy Extensions
705 == 1 : enable Privacy Extensions, but prefer public
706 addresses over temporary addresses.
707 > 1 : enable Privacy Extensions and prefer temporary
708 addresses over public addresses.
709 Default: 0 (for most devices)
710 -1 (for point-to-point devices and loopback devices)
712 temp_valid_lft - INTEGER
713 valid lifetime (in seconds) for temporary addresses.
714 Default: 604800 (7 days)
716 temp_prefered_lft - INTEGER
717 Preferred lifetime (in seconds) for temorary addresses.
718 Default: 86400 (1 day)
720 max_desync_factor - INTEGER
721 Maximum value for DESYNC_FACTOR, which is a random value
722 that ensures that clients don't synchronize with each
723 other and generage new addresses at exactly the same time.
727 regen_max_retry - INTEGER
728 Number of attempts before give up attempting to generate
729 valid temporary addresses.
732 max_addresses - INTEGER
733 Number of maximum addresses per interface. 0 disables limitation.
734 It is recommended not set too large value (or 0) because it would
735 be too easy way to crash kernel to allow to create too much of
736 autoconfigured addresses.
741 Limit the maximal rates for sending ICMPv6 packets.
742 0 to disable any limiting, otherwise the maximal rate in jiffies(1)
747 Pekka Savola <pekkas@netcore.fi>
748 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
751 /proc/sys/net/bridge/* Variables:
753 bridge-nf-call-arptables - BOOLEAN
754 1 : pass bridged ARP traffic to arptables' FORWARD chain.
758 bridge-nf-call-iptables - BOOLEAN
759 1 : pass bridged IPv4 traffic to iptables' chains.
763 bridge-nf-filter-vlan-tagged - BOOLEAN
764 1 : pass bridged vlan-tagged ARP/IP traffic to arptables/iptables.
769 $Id: ip-sysctl.txt,v 1.20 2001/12/13 09:00:18 davem Exp $