blob: a476b08a43e0dac5642a0bdfba76b8523be1033e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/proc/sys/net/ipv4/* Variables:
2
3ip_forward - BOOLEAN
4 0 - disabled (default)
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00005 not 0 - enabled
Linus Torvalds1da177e2005-04-16 15:20:36 -07006
7 Forward Packets between interfaces.
8
9 This variable is special, its change resets all configuration
10 parameters to their default state (RFC1122 for hosts, RFC1812
11 for routers)
12
13ip_default_ttl - INTEGER
Eric Dumazetcc6f02d2010-12-13 12:50:49 -080014 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)
Linus Torvalds1da177e2005-04-16 15:20:36 -070017
Hannes Frederic Sowacd174e62013-12-14 05:13:45 +010018ip_no_pmtu_disc - INTEGER
19 Disable Path MTU Discovery. If enabled in mode 1 and a
Hannes Frederic Sowa188b04d2013-12-14 04:42:13 +010020 fragmentation-required ICMP is received, the PMTU to this
21 destination will be set to min_pmtu (see below). You will need
22 to raise min_pmtu to the smallest interface MTU on your system
23 manually if you want to avoid locally generated fragments.
Hannes Frederic Sowacd174e62013-12-14 05:13:45 +010024
25 In mode 2 incoming Path MTU Discovery messages will be
26 discarded. Outgoing frames are handled the same as in mode 1,
27 implicitly setting IP_PMTUDISC_DONT on every created socket.
28
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +010029 Mode 3 is a hardend pmtu discover mode. The kernel will only
30 accept fragmentation-needed errors if the underlying protocol
31 can verify them besides a plain socket lookup. Current
32 protocols for which pmtu events will be honored are TCP, SCTP
33 and DCCP as they verify e.g. the sequence number or the
34 association. This mode should not be enabled globally but is
35 only intended to secure e.g. name servers in namespaces where
36 TCP path mtu must still work but path MTU information of other
37 protocols should be discarded. If enabled globally this mode
38 could break other protocols.
39
40 Possible values: 0-3
Hannes Frederic Sowa188b04d2013-12-14 04:42:13 +010041 Default: FALSE
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43min_pmtu - INTEGER
Eric Dumazet20db93c2011-11-08 14:21:44 -050044 default 552 - minimum discovered Path MTU
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Hannes Frederic Sowaf87c10a2014-01-09 10:01:15 +010046ip_forward_use_pmtu - BOOLEAN
47 By default we don't trust protocol path MTUs while forwarding
48 because they could be easily forged and can lead to unwanted
49 fragmentation by the router.
50 You only need to enable this if you have user-space software
51 which tries to discover path mtus by itself and depends on the
52 kernel honoring this information. This is normally not the
53 case.
54 Default: 0 (disabled)
55 Possible values:
56 0 - disabled
57 1 - enabled
58
Loganaden Velvindron219b5f22014-11-04 03:02:49 -080059fwmark_reflect - BOOLEAN
60 Controls the fwmark of kernel-generated IPv4 reply packets that are not
61 associated with a socket for example, TCP RSTs or ICMP echo replies).
62 If unset, these packets have a fwmark of zero. If set, they have the
63 fwmark of the packet they are replying to.
64 Default: 0
65
Ben Greearcbaf0872010-11-08 09:13:48 +000066route/max_size - INTEGER
67 Maximum number of routes allowed in the kernel. Increase
68 this when using large numbers of interfaces and/or routes.
69
YOSHIFUJI Hideaki / 吉藤英明27246802013-01-22 05:20:05 +000070neigh/default/gc_thresh1 - INTEGER
71 Minimum number of entries to keep. Garbage collector will not
72 purge entries if there are fewer than this number.
Li RongQingb66c66d2013-03-14 22:49:47 +000073 Default: 128
YOSHIFUJI Hideaki / 吉藤英明27246802013-01-22 05:20:05 +000074
stephen hemmingera3d12142014-08-25 15:05:30 -070075neigh/default/gc_thresh2 - INTEGER
76 Threshold when garbage collector becomes more aggressive about
77 purging entries. Entries older than 5 seconds will be cleared
78 when over this number.
79 Default: 512
80
Ben Greearcbaf0872010-11-08 09:13:48 +000081neigh/default/gc_thresh3 - INTEGER
82 Maximum number of neighbor entries allowed. Increase this
83 when using large numbers of interfaces and when communicating
84 with large numbers of directly-connected peers.
Shan Weicc868022012-12-04 18:50:35 +000085 Default: 1024
Ben Greearcbaf0872010-11-08 09:13:48 +000086
Eric Dumazet8b5c1712011-11-09 12:07:14 +000087neigh/default/unres_qlen_bytes - INTEGER
88 The maximum number of bytes which may be used by packets
89 queued for each unresolved address by other network layers.
90 (added in linux 3.3)
stephen hemminger3b09adc2013-01-03 07:50:29 +000091 Setting negative value is meaningless and will return error.
Shan Weicc868022012-12-04 18:50:35 +000092 Default: 65536 Bytes(64KB)
Eric Dumazet8b5c1712011-11-09 12:07:14 +000093
94neigh/default/unres_qlen - INTEGER
95 The maximum number of packets which may be queued for each
96 unresolved address by other network layers.
97 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
Shan Weicc868022012-12-04 18:50:35 +000098 Prior to linux 3.3, the default value is 3 which may cause
Shan Wei5d248c42012-12-06 16:27:51 +000099 unexpected packet loss. The current default value is calculated
Shan Weicc868022012-12-04 18:50:35 +0000100 according to default value of unres_qlen_bytes and true size of
101 packet.
102 Default: 31
Eric Dumazet8b5c1712011-11-09 12:07:14 +0000103
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104mtu_expires - INTEGER
105 Time, in seconds, that cached PMTU information is kept.
106
107min_adv_mss - INTEGER
108 The advertised MSS depends on the first hop route MTU, but will
109 never be lower than this setting.
110
111IP Fragmentation:
112
113ipfrag_high_thresh - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000114 Maximum memory used to reassemble IP fragments. When
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 ipfrag_high_thresh bytes of memory is allocated for this purpose,
116 the fragment handler will toss packets until ipfrag_low_thresh
Nikolay Aleksandrov1bab4c72014-07-24 16:50:37 +0200117 is reached. This also serves as a maximum limit to namespaces
118 different from the initial one.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000119
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120ipfrag_low_thresh - INTEGER
Florian Westphalb13d3cb2014-07-24 16:50:32 +0200121 Maximum memory used to reassemble IP fragments before the kernel
122 begins to remove incomplete fragment queues to free up resources.
123 The kernel still accepts new fragments for defragmentation.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125ipfrag_time - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000126 Time in seconds to keep an IP fragment in memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Herbert Xu89cee8b2005-12-13 23:14:27 -0800128ipfrag_max_dist - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000129 ipfrag_max_dist is a non-negative integer value which defines the
130 maximum "disorder" which is allowed among fragments which share a
131 common IP source address. Note that reordering of packets is
132 not unusual, but if a large number of fragments arrive from a source
133 IP address while a particular fragment queue remains incomplete, it
134 probably indicates that one or more fragments belonging to that queue
135 have been lost. When ipfrag_max_dist is positive, an additional check
136 is done on fragments before they are added to a reassembly queue - if
137 ipfrag_max_dist (or more) fragments have arrived from a particular IP
138 address between additions to any IP fragment queue using that source
139 address, it's presumed that one or more fragments in the queue are
140 lost. The existing fragment queue will be dropped, and a new one
Herbert Xu89cee8b2005-12-13 23:14:27 -0800141 started. An ipfrag_max_dist value of zero disables this check.
142
143 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
144 result in unnecessarily dropping fragment queues when normal
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000145 reordering of packets occurs, which could lead to poor application
146 performance. Using a very large value, e.g. 50000, increases the
147 likelihood of incorrectly reassembling IP fragments that originate
Herbert Xu89cee8b2005-12-13 23:14:27 -0800148 from different IP datagrams, which could result in data corruption.
149 Default: 64
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151INET peer storage:
152
153inet_peer_threshold - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000154 The approximate size of the storage. Starting from this threshold
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 entries will be thrown aggressively. This threshold also determines
156 entries' time-to-live and time intervals between garbage collection
157 passes. More entries, less time-to-live, less GC interval.
158
159inet_peer_minttl - INTEGER
160 Minimum time-to-live of entries. Should be enough to cover fragment
161 time-to-live on the reassembling side. This minimum time-to-live is
162 guaranteed if the pool size is less than inet_peer_threshold.
Stephen Hemminger77a538d2008-07-01 17:22:48 -0700163 Measured in seconds.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
165inet_peer_maxttl - INTEGER
166 Maximum time-to-live of entries. Unused entries will expire after
167 this period of time if there is no memory pressure on the pool (i.e.
168 when the number of entries in the pool is very small).
Stephen Hemminger77a538d2008-07-01 17:22:48 -0700169 Measured in seconds.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000171TCP variables:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800173somaxconn - INTEGER
174 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
175 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
176 for TCP sockets.
177
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800178tcp_abort_on_overflow - BOOLEAN
179 If listening service is too slow to accept new connections,
180 reset them. Default state is FALSE. It means that if overflow
181 occurred due to a burst, connection will recover. Enable this
182 option _only_ if you are really sure that listening daemon
183 cannot be tuned to accept connections faster. Enabling this
184 option can harm clients of your server.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800186tcp_adv_win_scale - INTEGER
187 Count buffering overhead as bytes/2^tcp_adv_win_scale
188 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
189 if it is <= 0.
Alexey Dobriyan0147fc02010-11-22 12:54:21 +0000190 Possible values are [-31, 31], inclusive.
Eric Dumazetb49960a2012-05-02 02:28:41 +0000191 Default: 1
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800192
193tcp_allowed_congestion_control - STRING
194 Show/set the congestion control choices available to non-privileged
195 processes. The list is a subset of those listed in
196 tcp_available_congestion_control.
197 Default is "reno" and the default setting (tcp_congestion_control).
198
199tcp_app_win - INTEGER
200 Reserve max(window/2^tcp_app_win, mss) of window for application
201 buffer. Value 0 is special, it means that nothing is reserved.
202 Default: 31
203
Eric Dumazetf54b3112013-12-05 22:36:05 -0800204tcp_autocorking - BOOLEAN
205 Enable TCP auto corking :
206 When applications do consecutive small write()/sendmsg() system calls,
207 we try to coalesce these small writes as much as possible, to lower
208 total amount of sent packets. This is done if at least one prior
209 packet for the flow is waiting in Qdisc queues or device transmit
210 queue. Applications can still use TCP_CORK for optimal behavior
211 when they know how/when to uncork their sockets.
212 Default : 1
213
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800214tcp_available_congestion_control - STRING
215 Shows the available congestion control choices that are registered.
216 More congestion control algorithms may be available as modules,
217 but not loaded.
218
John Heffner71599cd2007-02-27 10:03:56 -0800219tcp_base_mss - INTEGER
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700220 The initial value of search_low to be used by the packetization layer
221 Path MTU discovery (MTU probing). If MTU probing is enabled,
222 this is the initial MSS used by the connection.
John Heffner71599cd2007-02-27 10:03:56 -0800223
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800224tcp_congestion_control - STRING
225 Set the congestion control algorithm to be used for new
226 connections. The algorithm "reno" is always available, but
227 additional choices may be available based on kernel configuration.
228 Default is set as part of kernel configuration.
Eric Dumazetd8a6e652011-11-30 01:02:41 +0000229 For passive connections, the listener congestion control choice
230 is inherited.
231 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800232
233tcp_dsack - BOOLEAN
234 Allows TCP to send "duplicate" SACKs.
235
Yuchung Chengeed530b2012-05-02 13:30:03 +0000236tcp_early_retrans - INTEGER
237 Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
238 for triggering fast retransmit when the amount of outstanding data is
239 small and when no previously unsent data can be transmitted (such
Nandita Dukkipati6ba8a3b2013-03-11 10:00:43 +0000240 that limited transmit could be used). Also controls the use of
Masanari Iida3dd17ed2013-05-24 07:05:59 +0000241 Tail loss probe (TLP) that converts RTOs occurring due to tail
Nandita Dukkipati6ba8a3b2013-03-11 10:00:43 +0000242 losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
Yuchung Chengeed530b2012-05-02 13:30:03 +0000243 Possible values:
244 0 disables ER
245 1 enables ER
246 2 enables ER but delays fast recovery and fast retransmit
247 by a fourth of RTT. This mitigates connection falsely
248 recovers when network has a small degree of reordering
249 (less than 3 packets).
Nandita Dukkipati6ba8a3b2013-03-11 10:00:43 +0000250 3 enables delayed ER and TLP.
251 4 enables TLP only.
252 Default: 3
Yuchung Chengeed530b2012-05-02 13:30:03 +0000253
Peter Chubb34a6ef32011-02-02 15:39:58 -0800254tcp_ecn - INTEGER
Rick Jones7e3a2dc2012-11-28 09:53:10 +0000255 Control use of Explicit Congestion Notification (ECN) by TCP.
256 ECN is used only when both ends of the TCP connection indicate
257 support for it. This feature is useful in avoiding losses due
258 to congestion by allowing supporting routers to signal
259 congestion before having to drop packets.
Ilpo Järvinen255cac92009-05-04 11:07:36 -0700260 Possible values are:
Rick Jones7e3a2dc2012-11-28 09:53:10 +0000261 0 Disable ECN. Neither initiate nor accept ECN.
Vijay Subramanian3d55b322013-01-09 12:21:30 +0000262 1 Enable ECN when requested by incoming connections and
263 also request ECN on outgoing connection attempts.
264 2 Enable ECN when requested by incoming connections
Rick Jones7e3a2dc2012-11-28 09:53:10 +0000265 but do not request ECN on outgoing connections.
Ilpo Järvinen255cac92009-05-04 11:07:36 -0700266 Default: 2
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800267
268tcp_fack - BOOLEAN
269 Enable FACK congestion avoidance and fast retransmission.
270 The value is not used, if tcp_sack is not enabled.
271
272tcp_fin_timeout - INTEGER
Rick Jonesd825da22012-12-10 11:33:00 +0000273 The length of time an orphaned (no longer referenced by any
274 application) connection will remain in the FIN_WAIT_2 state
275 before it is aborted at the local end. While a perfectly
276 valid "receive only" state for an un-orphaned connection, an
277 orphaned connection in FIN_WAIT_2 state could otherwise wait
278 forever for the remote to close its end of the connection.
279 Cf. tcp_max_orphans
280 Default: 60 seconds
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800281
Ilpo Järvinen89808062007-02-27 10:10:55 -0800282tcp_frto - INTEGER
Yuchung Chenge33099f2013-03-20 13:33:00 +0000283 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
Ilpo Järvinencd998892007-09-20 11:35:26 -0700284 F-RTO is an enhanced recovery algorithm for TCP retransmission
Yuchung Chenge33099f2013-03-20 13:33:00 +0000285 timeouts. It is particularly beneficial in networks where the
286 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
287 modification. It does not require any support from the peer.
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700288
Yuchung Chenge33099f2013-03-20 13:33:00 +0000289 By default it's enabled with a non-zero value. 0 disables F-RTO.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290
291tcp_keepalive_time - INTEGER
292 How often TCP sends out keepalive messages when keepalive is enabled.
293 Default: 2hours.
294
295tcp_keepalive_probes - INTEGER
296 How many keepalive probes TCP sends out, until it decides that the
297 connection is broken. Default value: 9.
298
299tcp_keepalive_intvl - INTEGER
300 How frequently the probes are send out. Multiplied by
301 tcp_keepalive_probes it is time to kill not responding connection,
302 after probes started. Default value: 75sec i.e. connection
303 will be aborted after ~11 minutes of retries.
304
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800305tcp_low_latency - BOOLEAN
306 If set, the TCP stack makes decisions that prefer lower
307 latency as opposed to higher throughput. By default, this
308 option is not set meaning that higher throughput is preferred.
309 An example of an application where this default should be
310 changed would be a Beowulf compute cluster.
311 Default: 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
313tcp_max_orphans - INTEGER
314 Maximal number of TCP sockets not attached to any user file handle,
315 held by system. If this number is exceeded orphaned connections are
316 reset immediately and warning is printed. This limit exists
317 only to prevent simple DoS attacks, you _must_ not rely on this
318 or lower the limit artificially, but rather increase it
319 (probably, after increasing installed memory),
320 if network conditions require more than default value,
321 and tune network services to linger and kill such states
322 more aggressively. Let me to remind again: each orphan eats
323 up to ~64K of unswappable memory.
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325tcp_max_syn_backlog - INTEGER
Peter Pan(潘卫平)99b53bd2011-12-05 21:39:41 +0000326 Maximal number of remembered connection requests, which have not
327 received an acknowledgment from connecting client.
328 The minimal value is 128 for low memory machines, and it will
329 increase in proportion to the memory of machine.
330 If server suffers from overload, try increasing this number.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800332tcp_max_tw_buckets - INTEGER
333 Maximal number of timewait sockets held by system simultaneously.
334 If this number is exceeded time-wait socket is immediately destroyed
335 and warning is printed. This limit exists only to prevent
336 simple DoS attacks, you _must_ not lower the limit artificially,
337 but rather increase it (probably, after increasing installed memory),
338 if network conditions require more than default value.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800340tcp_mem - vector of 3 INTEGERs: min, pressure, max
341 min: below this number of pages TCP is not bothered about its
342 memory appetite.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800344 pressure: when amount of memory allocated by TCP exceeds this number
345 of pages, TCP moderates its memory consumption and enters memory
346 pressure mode, which is exited when memory consumption falls
347 under "min".
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800349 max: number of pages allowed for queueing by all TCP sockets.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800351 Defaults are calculated at boot time from amount of available
352 memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
John Heffner71599cd2007-02-27 10:03:56 -0800354tcp_moderate_rcvbuf - BOOLEAN
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700355 If set, TCP performs receive buffer auto-tuning, attempting to
John Heffner71599cd2007-02-27 10:03:56 -0800356 automatically size the buffer (no greater than tcp_rmem[2]) to
357 match the size required by the path for full throughput. Enabled by
358 default.
359
360tcp_mtu_probing - INTEGER
361 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
362 values:
363 0 - Disabled
364 1 - Disabled by default, enabled when an ICMP black hole detected
365 2 - Always enabled, use initial MSS of tcp_base_mss.
366
367tcp_no_metrics_save - BOOLEAN
368 By default, TCP saves various connection metrics in the route cache
369 when the connection closes, so that connections established in the
370 near future can use these to set initial conditions. Usually, this
371 increases overall performance, but may sometimes cause performance
Simon Arlott0f035b82007-10-20 01:30:25 +0200372 degradation. If set, TCP will not cache metrics on closing
John Heffner71599cd2007-02-27 10:03:56 -0800373 connections.
374
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800375tcp_orphan_retries - INTEGER
Damian Lukowski5d789222009-09-01 10:24:04 +0000376 This value influences the timeout of a locally closed TCP connection,
377 when RTO retransmissions remain unacknowledged.
378 See tcp_retries2 for more details.
379
David S. Miller06b8fc52011-07-08 09:31:31 -0700380 The default value is 8.
Damian Lukowski5d789222009-09-01 10:24:04 +0000381 If your machine is a loaded WEB server,
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800382 you should think about lowering this value, such sockets
383 may consume significant resources. Cf. tcp_max_orphans.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
385tcp_reordering - INTEGER
386 Maximal reordering of packets in a TCP stream.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000387 Default: 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389tcp_retrans_collapse - BOOLEAN
390 Bug-to-bug compatibility with some broken printers.
391 On retransmit try to send bigger packets to work around bugs in
392 certain TCP stacks.
393
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800394tcp_retries1 - INTEGER
Damian Lukowski5d789222009-09-01 10:24:04 +0000395 This value influences the time, after which TCP decides, that
396 something is wrong due to unacknowledged RTO retransmissions,
397 and reports this suspicion to the network layer.
398 See tcp_retries2 for more details.
399
400 RFC 1122 recommends at least 3 retransmissions, which is the
401 default.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800403tcp_retries2 - INTEGER
Damian Lukowski5d789222009-09-01 10:24:04 +0000404 This value influences the timeout of an alive TCP connection,
405 when RTO retransmissions remain unacknowledged.
406 Given a value of N, a hypothetical TCP connection following
407 exponential backoff with an initial RTO of TCP_RTO_MIN would
408 retransmit N times before killing the connection at the (N+1)th RTO.
409
410 The default value of 15 yields a hypothetical timeout of 924.6
411 seconds and is a lower bound for the effective timeout.
412 TCP will effectively time out at the first RTO which exceeds the
413 hypothetical timeout.
414
415 RFC 1122 recommends at least 100 seconds for the timeout,
416 which corresponds to a value of at least 8.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800418tcp_rfc1337 - BOOLEAN
419 If set, the TCP stack behaves conforming to RFC1337. If unset,
420 we are not conforming to RFC, but prevent TCP TIME_WAIT
421 assassination.
422 Default: 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
424tcp_rmem - vector of 3 INTEGERs: min, default, max
425 min: Minimal size of receive buffer used by TCP sockets.
426 It is guaranteed to each TCP socket, even under moderate memory
427 pressure.
Max Matveev6539fef2011-06-21 21:18:13 +0000428 Default: 1 page
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
J. Bruce Fields53025f52008-07-10 16:47:41 -0700430 default: initial size of receive buffer used by TCP sockets.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 This value overrides net.core.rmem_default used by other protocols.
432 Default: 87380 bytes. This value results in window of 65535 with
433 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
434 less for default tcp_app_win. See below about these variables.
435
436 max: maximal size of receive buffer allowed for automatically
437 selected receiver buffers for TCP socket. This value does not override
J. Bruce Fields53025f52008-07-10 16:47:41 -0700438 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
439 automatic tuning of that socket's receive buffer size, in which
440 case this value is ignored.
Eric Dumazetb49960a2012-05-02 02:28:41 +0000441 Default: between 87380B and 6MB, depending on RAM size.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800443tcp_sack - BOOLEAN
444 Enable select acknowledgments (SACKS).
Rick Jones15d99e02006-03-20 22:40:29 -0800445
David S. Miller35089bb2006-06-13 22:33:04 -0700446tcp_slow_start_after_idle - BOOLEAN
447 If set, provide RFC2861 behavior and time out the congestion
448 window after an idle period. An idle period is defined at
449 the current RTO. If unset, the congestion window will not
450 be timed out after an idle period.
451 Default: 1
452
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800453tcp_stdurg - BOOLEAN
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700454 Use the Host requirements interpretation of the TCP urgent pointer field.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800455 Most hosts use the older BSD interpretation, so if you turn this on
456 Linux might not communicate correctly with them.
457 Default: FALSE
458
459tcp_synack_retries - INTEGER
460 Number of times SYNACKs for a passive TCP connection attempt will
461 be retransmitted. Should not be higher than 255. Default value
Alex Bergmann6c9ff972012-08-31 02:48:31 +0000462 is 5, which corresponds to 31seconds till the last retransmission
463 with the current initial RTO of 1second. With this the final timeout
464 for a passive TCP connection will happen after 63seconds.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800465
466tcp_syncookies - BOOLEAN
Shan Weia3c910d2013-06-21 15:18:32 +0800467 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800468 Send out syncookies when the syn backlog queue of a socket
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700469 overflows. This is to prevent against the common 'SYN flood attack'
Shan Weia3c910d2013-06-21 15:18:32 +0800470 Default: 1
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800471
472 Note, that syncookies is fallback facility.
473 It MUST NOT be used to help highly loaded servers to stand
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700474 against legal connection rate. If you see SYN flood warnings
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800475 in your logs, but investigation shows that they occur
476 because of overload with legal connections, you should tune
477 another parameters until this warning disappear.
478 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
479
480 syncookies seriously violate TCP protocol, do not allow
481 to use TCP extensions, can result in serious degradation
482 of some services (f.e. SMTP relaying), visible not by you,
483 but your clients and relays, contacting you. While you see
Stephen Hemminger4edc2f32008-07-10 16:50:26 -0700484 SYN flood warnings in logs not being really flooded, your server
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800485 is seriously misconfigured.
486
Hannes Frederic Sowa5ad37d52013-07-26 17:43:23 +0200487 If you want to test which effects syncookies have to your
488 network connections you can set this knob to 2 to enable
489 unconditionally generation of syncookies.
490
Yuchung Chengcf60af02012-07-19 06:43:09 +0000491tcp_fastopen - INTEGER
492 Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
493 in the opening SYN packet. To use this feature, the client application
Jerry Chu10467162012-08-31 12:29:11 +0000494 must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
495 connect() to perform a TCP handshake automatically.
Yuchung Chengcf60af02012-07-19 06:43:09 +0000496
Jerry Chu10467162012-08-31 12:29:11 +0000497 The values (bitmap) are
Yuchung Cheng0d41cca2013-10-31 09:19:32 -0700498 1: Enables sending data in the opening SYN on the client w/ MSG_FASTOPEN.
Jerry Chu10467162012-08-31 12:29:11 +0000499 2: Enables TCP Fast Open on the server side, i.e., allowing data in
500 a SYN packet to be accepted and passed to the application before
501 3-way hand shake finishes.
502 4: Send data in the opening SYN regardless of cookie availability and
503 without a cookie option.
504 0x100: Accept SYN data w/o validating the cookie.
505 0x200: Accept data-in-SYN w/o any cookie option present.
506 0x400/0x800: Enable Fast Open on all listeners regardless of the
507 TCP_FASTOPEN socket option. The two different flags designate two
508 different ways of setting max_qlen without the TCP_FASTOPEN socket
509 option.
Yuchung Chengcf60af02012-07-19 06:43:09 +0000510
Yuchung Cheng0d41cca2013-10-31 09:19:32 -0700511 Default: 1
Yuchung Chengcf60af02012-07-19 06:43:09 +0000512
Jerry Chu10467162012-08-31 12:29:11 +0000513 Note that the client & server side Fast Open flags (1 and 2
514 respectively) must be also enabled before the rest of flags can take
515 effect.
516
517 See include/net/tcp.h and the code for more details.
518
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800519tcp_syn_retries - INTEGER
520 Number of times initial SYNs for an active TCP connection attempt
521 will be retransmitted. Should not be higher than 255. Default value
stephen hemminger3b09adc2013-01-03 07:50:29 +0000522 is 6, which corresponds to 63seconds till the last retransmission
Alex Bergmann6c9ff972012-08-31 02:48:31 +0000523 with the current initial RTO of 1second. With this the final timeout
524 for an active TCP connection attempt will happen after 127seconds.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800525
526tcp_timestamps - BOOLEAN
527 Enable timestamps as defined in RFC1323.
528
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700529tcp_min_tso_segs - INTEGER
530 Minimal number of segments per TSO frame.
531 Since linux-3.12, TCP does an automatic sizing of TSO frames,
532 depending on flow rate, instead of filling 64Kbytes packets.
533 For specific usages, it's possible to force TCP to build big
534 TSO frames. Note that TCP stack might split too big TSO packets
535 if available window is too small.
536 Default: 2
537
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800538tcp_tso_win_divisor - INTEGER
539 This allows control over what percentage of the congestion window
540 can be consumed by a single TSO frame.
541 The setting of this parameter is a choice between burstiness and
542 building larger TSO frames.
543 Default: 3
544
545tcp_tw_recycle - BOOLEAN
546 Enable fast recycling TIME-WAIT sockets. Default value is 0.
547 It should not be changed without advice/request of technical
548 experts.
549
550tcp_tw_reuse - BOOLEAN
551 Allow to reuse TIME-WAIT sockets for new connections when it is
552 safe from protocol viewpoint. Default value is 0.
553 It should not be changed without advice/request of technical
554 experts.
555
556tcp_window_scaling - BOOLEAN
557 Enable window scaling as defined in RFC1323.
558
559tcp_wmem - vector of 3 INTEGERs: min, default, max
J. Bruce Fields53025f52008-07-10 16:47:41 -0700560 min: Amount of memory reserved for send buffers for TCP sockets.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800561 Each TCP socket has rights to use it due to fact of its birth.
Max Matveev6539fef2011-06-21 21:18:13 +0000562 Default: 1 page
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800563
J. Bruce Fields53025f52008-07-10 16:47:41 -0700564 default: initial size of send buffer used by TCP sockets. This
565 value overrides net.core.wmem_default used by other protocols.
566 It is usually lower than net.core.wmem_default.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800567 Default: 16K
568
J. Bruce Fields53025f52008-07-10 16:47:41 -0700569 max: Maximal amount of memory allowed for automatically tuned
570 send buffers for TCP sockets. This value does not override
571 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
572 automatic tuning of that socket's send buffer size, in which case
573 this value is ignored.
574 Default: between 64K and 4MB, depending on RAM size.
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800575
Eric Dumazetc9bee3b72013-07-22 20:27:07 -0700576tcp_notsent_lowat - UNSIGNED INTEGER
577 A TCP socket can control the amount of unsent bytes in its write queue,
578 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
579 reports POLLOUT events if the amount of unsent bytes is below a per
580 socket value, and if the write queue is not full. sendmsg() will
581 also not add new buffers if the limit is hit.
582
583 This global variable controls the amount of unsent data for
584 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
585 to the global variable has immediate effect.
586
587 Default: UINT_MAX (0xFFFFFFFF)
588
Stephen Hemmingeref56e622006-11-09 16:37:26 -0800589tcp_workaround_signed_windows - BOOLEAN
590 If set, assume no receipt of a window scaling option means the
591 remote TCP is broken and treats the window as a signed quantity.
592 If unset, assume the remote TCP is not broken even if we do
593 not receive a window scaling option from them.
594 Default: 0
595
Andreas Petlund36e31b0a2010-02-18 02:47:01 +0000596tcp_thin_linear_timeouts - BOOLEAN
597 Enable dynamic triggering of linear timeouts for thin streams.
598 If set, a check is performed upon retransmission by timeout to
599 determine if the stream is thin (less than 4 packets in flight).
600 As long as the stream is found to be thin, up to 6 linear
601 timeouts may be performed before exponential backoff mode is
602 initiated. This improves retransmission latency for
603 non-aggressive thin streams, often found to be time-dependent.
604 For more information on thin streams, see
605 Documentation/networking/tcp-thin.txt
606 Default: 0
607
Andreas Petlund7e380172010-02-18 04:48:19 +0000608tcp_thin_dupack - BOOLEAN
609 Enable dynamic triggering of retransmissions after one dupACK
610 for thin streams. If set, a check is performed upon reception
611 of a dupACK to determine if the stream is thin (less than 4
612 packets in flight). As long as the stream is found to be thin,
613 data is retransmitted on the first received dupACK. This
614 improves retransmission latency for non-aggressive thin
615 streams, often found to be time-dependent.
616 For more information on thin streams, see
617 Documentation/networking/tcp-thin.txt
618 Default: 0
619
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000620tcp_limit_output_bytes - INTEGER
621 Controls TCP Small Queue limit per tcp socket.
622 TCP bulk sender tends to increase packets in flight until it
623 gets losses notifications. With SNDBUF autotuning, this can
624 result in a large amount of packets queued in qdisc/device
625 on the local machine, hurting latency of other flows, for
626 typical pfifo_fast qdiscs.
627 tcp_limit_output_bytes limits the number of bytes on qdisc
628 or device to reduce artificial RTT/cwnd and reduce bufferbloat.
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000629 Default: 131072
630
Eric Dumazet282f23c2012-07-17 10:13:05 +0200631tcp_challenge_ack_limit - INTEGER
632 Limits number of Challenge ACK sent per second, as recommended
633 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
634 Default: 100
635
Hideo Aoki95766ff2007-12-31 00:29:24 -0800636UDP variables:
637
638udp_mem - vector of 3 INTEGERs: min, pressure, max
639 Number of pages allowed for queueing by all UDP sockets.
640
641 min: Below this number of pages UDP is not bothered about its
642 memory appetite. When amount of memory allocated by UDP exceeds
643 this number, UDP starts to moderate memory usage.
644
645 pressure: This value was introduced to follow format of tcp_mem.
646
647 max: Number of pages allowed for queueing by all UDP sockets.
648
649 Default is calculated at boot time from amount of available memory.
650
651udp_rmem_min - INTEGER
652 Minimal size of receive buffer used by UDP sockets in moderation.
653 Each UDP socket is able to use the size for receiving data, even if
654 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
Max Matveev6539fef2011-06-21 21:18:13 +0000655 Default: 1 page
Hideo Aoki95766ff2007-12-31 00:29:24 -0800656
657udp_wmem_min - INTEGER
658 Minimal size of send buffer used by UDP sockets in moderation.
659 Each UDP socket is able to use the size for sending data, even if
660 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
Max Matveev6539fef2011-06-21 21:18:13 +0000661 Default: 1 page
Hideo Aoki95766ff2007-12-31 00:29:24 -0800662
Paul Moore8802f612006-08-03 16:45:49 -0700663CIPSOv4 Variables:
664
665cipso_cache_enable - BOOLEAN
666 If set, enable additions to and lookups from the CIPSO label mapping
667 cache. If unset, additions are ignored and lookups always result in a
668 miss. However, regardless of the setting the cache is still
669 invalidated when required when means you can safely toggle this on and
670 off and the cache will always be "safe".
671 Default: 1
672
673cipso_cache_bucket_size - INTEGER
674 The CIPSO label cache consists of a fixed size hash table with each
675 hash bucket containing a number of cache entries. This variable limits
676 the number of entries in each hash bucket; the larger the value the
677 more CIPSO label mappings that can be cached. When the number of
678 entries in a given hash bucket reaches this limit adding new entries
679 causes the oldest entry in the bucket to be removed to make room.
680 Default: 10
681
682cipso_rbm_optfmt - BOOLEAN
683 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
684 the CIPSO draft specification (see Documentation/netlabel for details).
685 This means that when set the CIPSO tag will be padded with empty
686 categories in order to make the packet data 32-bit aligned.
687 Default: 0
688
689cipso_rbm_structvalid - BOOLEAN
690 If set, do a very strict check of the CIPSO option when
691 ip_options_compile() is called. If unset, relax the checks done during
692 ip_options_compile(). Either way is "safe" as errors are caught else
693 where in the CIPSO processing code but setting this to 0 (False) should
694 result in less work (i.e. it should be faster) but could cause problems
695 with other implementations that require strict checking.
696 Default: 0
697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698IP Variables:
699
700ip_local_port_range - 2 INTEGERS
701 Defines the local port range that is used by TCP and UDP to
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000702 choose the local port. The first number is the first, the
Fernando Luis Vazquez Cao5d6bd862012-04-03 08:41:40 +0000703 second the last local port number. The default values are
704 32768 and 61000 respectively.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
Amerigo Wange3826f12010-05-05 00:27:06 +0000706ip_local_reserved_ports - list of comma separated ranges
707 Specify the ports which are reserved for known third-party
708 applications. These ports will not be used by automatic port
709 assignments (e.g. when calling connect() or bind() with port
710 number 0). Explicit port allocation behavior is unchanged.
711
712 The format used for both input and output is a comma separated
713 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
714 10). Writing to the file will clear all previously reserved
715 ports and update the current list with the one given in the
716 input.
717
718 Note that ip_local_port_range and ip_local_reserved_ports
719 settings are independent and both are considered by the kernel
720 when determining which ports are available for automatic port
721 assignments.
722
723 You can reserve ports which are not in the current
724 ip_local_port_range, e.g.:
725
726 $ cat /proc/sys/net/ipv4/ip_local_port_range
727 32000 61000
728 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
729 8080,9148
730
731 although this is redundant. However such a setting is useful
732 if later the port range is changed to a value that will
733 include the reserved ports.
734
735 Default: Empty
736
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737ip_nonlocal_bind - BOOLEAN
738 If set, allows processes to bind() to non-local IP addresses,
739 which can be quite useful - but may break some applications.
740 Default: 0
741
742ip_dynaddr - BOOLEAN
743 If set non-zero, enables support for dynamic addresses.
744 If set to a non-zero value larger than 1, a kernel log
745 message will be printed when dynamic address rewriting
746 occurs.
747 Default: 0
748
Cong Wange3d73bc2013-06-11 18:54:39 +0800749ip_early_demux - BOOLEAN
750 Optimize input packet processing down to one demux for
751 certain kinds of local sockets. Currently we only do this
752 for established TCP sockets.
753
754 It may add an additional cost for pure routing workloads that
755 reduces overall throughput, in such case you should disable it.
756 Default: 1
757
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758icmp_echo_ignore_all - BOOLEAN
David S. Miller7ce312462005-10-03 16:07:30 -0700759 If set non-zero, then the kernel will ignore all ICMP ECHO
760 requests sent to it.
761 Default: 0
762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763icmp_echo_ignore_broadcasts - BOOLEAN
David S. Miller7ce312462005-10-03 16:07:30 -0700764 If set non-zero, then the kernel will ignore all ICMP ECHO and
765 TIMESTAMP requests sent to it via broadcast/multicast.
766 Default: 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
768icmp_ratelimit - INTEGER
769 Limit the maximal rates for sending ICMP packets whose type matches
770 icmp_ratemask (see below) to specific targets.
Stephen Hemminger6dbf4bc2008-07-01 19:29:07 -0700771 0 to disable any limiting,
772 otherwise the minimal space between responses in milliseconds.
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700773 Note that another sysctl, icmp_msgs_per_sec limits the number
774 of ICMP packets sent on all targets.
Stephen Hemminger6dbf4bc2008-07-01 19:29:07 -0700775 Default: 1000
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Eric Dumazet4cdf5072014-09-19 07:38:40 -0700777icmp_msgs_per_sec - INTEGER
778 Limit maximal number of ICMP packets sent per second from this host.
779 Only messages whose type matches icmp_ratemask (see below) are
780 controlled by this limit.
781 Default: 1000
782
783icmp_msgs_burst - INTEGER
784 icmp_msgs_per_sec controls number of ICMP packets sent per second,
785 while icmp_msgs_burst controls the burst size of these packets.
786 Default: 50
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788icmp_ratemask - INTEGER
789 Mask made of ICMP types for which rates are being limited.
790 Significant bits: IHGFEDCBA9876543210
791 Default mask: 0000001100000011000 (6168)
792
793 Bit definitions (see include/linux/icmp.h):
794 0 Echo Reply
795 3 Destination Unreachable *
796 4 Source Quench *
797 5 Redirect
798 8 Echo Request
799 B Time Exceeded *
800 C Parameter Problem *
801 D Timestamp Request
802 E Timestamp Reply
803 F Info Request
804 G Info Reply
805 H Address Mask Request
806 I Address Mask Reply
807
808 * These are rate limited by default (see default mask above)
809
810icmp_ignore_bogus_error_responses - BOOLEAN
811 Some routers violate RFC1122 by sending bogus responses to broadcast
812 frames. Such violations are normally logged via a kernel warning.
813 If this is set to TRUE, the kernel will not give such warnings, which
814 will avoid log file clutter.
Rami Rosene8b265e2013-06-07 20:16:19 +0000815 Default: 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816
Horms95f7daf2006-02-02 17:02:25 -0800817icmp_errors_use_inbound_ifaddr - BOOLEAN
818
819 If zero, icmp error messages are sent with the primary address of
820 the exiting interface.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000821
Horms95f7daf2006-02-02 17:02:25 -0800822 If non-zero, the message will be sent with the primary address of
823 the interface that received the packet that caused the icmp error.
824 This is the behaviour network many administrators will expect from
825 a router. And it can make debugging complicated network layouts
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000826 much easier.
Horms95f7daf2006-02-02 17:02:25 -0800827
828 Note that if no primary address exists for the interface selected,
829 then the primary address of the first non-loopback interface that
Matt LaPlanted6bc8ac2006-10-03 22:54:15 +0200830 has one will be used regardless of this setting.
Horms95f7daf2006-02-02 17:02:25 -0800831
832 Default: 0
833
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834igmp_max_memberships - INTEGER
835 Change the maximum number of multicast groups we can subscribe to.
836 Default: 20
837
Jeremy Ederd67ef352010-11-15 05:41:31 +0000838 Theoretical maximum value is bounded by having to send a membership
839 report in a single datagram (i.e. the report can't span multiple
840 datagrams, or risk confusing the switch and leaving groups you don't
841 intend to).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842
Jeremy Ederd67ef352010-11-15 05:41:31 +0000843 The number of supported groups 'M' is bounded by the number of group
844 report entries you can fit into a single datagram of 65535 bytes.
845
846 M = 65536-sizeof (ip header)/(sizeof(Group record))
847
848 Group records are variable length, with a minimum of 12 bytes.
849 So net.ipv4.igmp_max_memberships should not be set higher than:
850
851 (65536-24) / 12 = 5459
852
853 The value 5459 assumes no IP header options, so in practice
854 this number may be lower.
855
856 conf/interface/* changes special settings per interface (where
857 "interface" is the name of your network interface)
858
859 conf/all/* is special, changes the settings for all interfaces
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
Hannes Frederic Sowaa9fe8e22014-09-02 15:49:26 +0200861igmp_qrv - INTEGER
862 Controls the IGMP query robustness variable (see RFC2236 8.1).
863 Default: 2 (as specified by RFC2236 8.1)
864 Minimum: 1 (as specified by RFC6636 4.5)
865
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866log_martians - BOOLEAN
867 Log packets with impossible addresses to kernel log.
868 log_martians for the interface will be enabled if at least one of
869 conf/{all,interface}/log_martians is set to TRUE,
870 it will be disabled otherwise
871
872accept_redirects - BOOLEAN
873 Accept ICMP redirect messages.
874 accept_redirects for the interface will be enabled if:
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000875 - both conf/{all,interface}/accept_redirects are TRUE in the case
876 forwarding for the interface is enabled
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 or
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000878 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
879 case forwarding for the interface is disabled
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 accept_redirects for the interface will be disabled otherwise
881 default TRUE (host)
882 FALSE (router)
883
884forwarding - BOOLEAN
885 Enable IP forwarding on this interface.
886
887mc_forwarding - BOOLEAN
888 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
889 and a multicast routing daemon is required.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000890 conf/all/mc_forwarding must also be set to TRUE to enable multicast
891 routing for the interface
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
893medium_id - INTEGER
894 Integer value used to differentiate the devices by the medium they
895 are attached to. Two devices can have different id values when
896 the broadcast packets are received only on one of them.
897 The default value 0 means that the device is the only interface
898 to its medium, value of -1 means that medium is not known.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000899
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 Currently, it is used to change the proxy_arp behavior:
901 the proxy_arp feature is enabled for packets forwarded between
902 two devices attached to different media.
903
904proxy_arp - BOOLEAN
905 Do proxy arp.
906 proxy_arp for the interface will be enabled if at least one of
907 conf/{all,interface}/proxy_arp is set to TRUE,
908 it will be disabled otherwise
909
Jesper Dangaard Brouer65324142010-01-05 05:50:47 +0000910proxy_arp_pvlan - BOOLEAN
911 Private VLAN proxy arp.
912 Basically allow proxy arp replies back to the same interface
913 (from which the ARP request/solicitation was received).
914
915 This is done to support (ethernet) switch features, like RFC
916 3069, where the individual ports are NOT allowed to
917 communicate with each other, but they are allowed to talk to
918 the upstream router. As described in RFC 3069, it is possible
919 to allow these hosts to communicate through the upstream
920 router by proxy_arp'ing. Don't need to be used together with
921 proxy_arp.
922
923 This technology is known by different names:
924 In RFC 3069 it is called VLAN Aggregation.
925 Cisco and Allied Telesyn call it Private VLAN.
926 Hewlett-Packard call it Source-Port filtering or port-isolation.
927 Ericsson call it MAC-Forced Forwarding (RFC Draft).
928
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929shared_media - BOOLEAN
930 Send(router) or accept(host) RFC1620 shared media redirects.
931 Overrides ip_secure_redirects.
932 shared_media for the interface will be enabled if at least one of
933 conf/{all,interface}/shared_media is set to TRUE,
934 it will be disabled otherwise
935 default TRUE
936
937secure_redirects - BOOLEAN
938 Accept ICMP redirect messages only for gateways,
939 listed in default gateway list.
940 secure_redirects for the interface will be enabled if at least one of
941 conf/{all,interface}/secure_redirects is set to TRUE,
942 it will be disabled otherwise
943 default TRUE
944
945send_redirects - BOOLEAN
946 Send redirects, if router.
947 send_redirects for the interface will be enabled if at least one of
948 conf/{all,interface}/send_redirects is set to TRUE,
949 it will be disabled otherwise
950 Default: TRUE
951
952bootp_relay - BOOLEAN
953 Accept packets with source address 0.b.c.d destined
954 not to this host as local ones. It is supposed, that
955 BOOTP relay daemon will catch and forward such packets.
956 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
957 for the interface
958 default FALSE
959 Not Implemented Yet.
960
961accept_source_route - BOOLEAN
962 Accept packets with SRR option.
963 conf/all/accept_source_route must also be set to TRUE to accept packets
964 with SRR option on the interface
965 default TRUE (router)
966 FALSE (host)
967
Patrick McHardy8153a102009-12-03 01:25:58 +0000968accept_local - BOOLEAN
Sébastien Barré72b126a2014-09-10 18:20:23 +0200969 Accept packets with local source addresses. In combination with
970 suitable routing, this can be used to direct packets between two
971 local interfaces over the wire and have them accepted properly.
Patrick McHardy8153a102009-12-03 01:25:58 +0000972 default FALSE
973
Thomas Grafd0daebc32012-06-12 00:44:01 +0000974route_localnet - BOOLEAN
975 Do not consider loopback addresses as martian source or destination
976 while routing. This enables the use of 127/8 for local routing purposes.
977 default FALSE
978
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000979rp_filter - INTEGER
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 0 - No source validation.
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000981 1 - Strict mode as defined in RFC3704 Strict Reverse Path
982 Each incoming packet is tested against the FIB and if the interface
983 is not the best reverse path the packet check will fail.
984 By default failed packets are discarded.
985 2 - Loose mode as defined in RFC3704 Loose Reverse Path
986 Each incoming packet's source address is also tested against the FIB
987 and if the source address is not reachable via any interface
988 the packet check will fail.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000990 Current recommended practice in RFC3704 is to enable strict mode
Jesper Dangaard Brouerbf869c32009-02-23 04:37:55 +0000991 to prevent IP spoofing from DDos attacks. If using asymmetric routing
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +0000992 or other complicated routing, then loose mode is recommended.
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000993
Shan Wei1f5865e2009-12-02 15:39:04 -0800994 The max value from conf/{all,interface}/rp_filter is used
995 when doing source validation on the {interface}.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996
997 Default value is 0. Note that some distributions enable it
998 in startup scripts.
999
1000arp_filter - BOOLEAN
1001 1 - Allows you to have multiple network interfaces on the same
1002 subnet, and have the ARPs for each interface be answered
1003 based on whether or not the kernel would route a packet from
1004 the ARP'd IP out that interface (therefore you must use source
1005 based routing for this to work). In other words it allows control
1006 of which cards (usually 1) will respond to an arp request.
1007
1008 0 - (default) The kernel can respond to arp requests with addresses
1009 from other interfaces. This may seem wrong but it usually makes
1010 sense, because it increases the chance of successful communication.
1011 IP addresses are owned by the complete host on Linux, not by
1012 particular interfaces. Only for more complex setups like load-
1013 balancing, does this behaviour cause problems.
1014
1015 arp_filter for the interface will be enabled if at least one of
1016 conf/{all,interface}/arp_filter is set to TRUE,
1017 it will be disabled otherwise
1018
1019arp_announce - INTEGER
1020 Define different restriction levels for announcing the local
1021 source IP address from IP packets in ARP requests sent on
1022 interface:
1023 0 - (default) Use any local address, configured on any interface
1024 1 - Try to avoid local addresses that are not in the target's
1025 subnet for this interface. This mode is useful when target
1026 hosts reachable via this interface require the source IP
1027 address in ARP requests to be part of their logical network
1028 configured on the receiving interface. When we generate the
1029 request we will check all our subnets that include the
1030 target IP and will preserve the source address if it is from
1031 such subnet. If there is no such subnet we select source
1032 address according to the rules for level 2.
1033 2 - Always use the best local address for this target.
1034 In this mode we ignore the source address in the IP packet
1035 and try to select local address that we prefer for talks with
1036 the target host. Such local address is selected by looking
1037 for primary IP addresses on all our subnets on the outgoing
1038 interface that include the target IP address. If no suitable
1039 local address is found we select the first local address
1040 we have on the outgoing interface or on all other interfaces,
1041 with the hope we will receive reply for our request and
1042 even sometimes no matter the source IP address we announce.
1043
1044 The max value from conf/{all,interface}/arp_announce is used.
1045
1046 Increasing the restriction level gives more chance for
1047 receiving answer from the resolved target while decreasing
1048 the level announces more valid sender's information.
1049
1050arp_ignore - INTEGER
1051 Define different modes for sending replies in response to
1052 received ARP requests that resolve local target IP addresses:
1053 0 - (default): reply for any local target IP address, configured
1054 on any interface
1055 1 - reply only if the target IP address is local address
1056 configured on the incoming interface
1057 2 - reply only if the target IP address is local address
1058 configured on the incoming interface and both with the
1059 sender's IP address are part from same subnet on this interface
1060 3 - do not reply for local addresses configured with scope host,
1061 only resolutions for global and link addresses are replied
1062 4-7 - reserved
1063 8 - do not reply for all local addresses
1064
1065 The max value from conf/{all,interface}/arp_ignore is used
1066 when ARP request is received on the {interface}
1067
Stephen Hemmingereefef1c2009-02-01 01:04:33 -08001068arp_notify - BOOLEAN
1069 Define mode for notification of address and device changes.
1070 0 - (default): do nothing
Ian Campbell3f8dc232010-05-26 00:09:41 +00001071 1 - Generate gratuitous arp requests when device is brought up
Stephen Hemmingereefef1c2009-02-01 01:04:33 -08001072 or hardware address changes.
1073
Neil Hormanc1b1bce2006-03-20 22:40:03 -08001074arp_accept - BOOLEAN
Octavian Purdila6d955182010-01-18 12:58:44 +00001075 Define behavior for gratuitous ARP frames who's IP is not
1076 already present in the ARP table:
1077 0 - don't create new entries in the ARP table
1078 1 - create new entries in the ARP table
1079
1080 Both replies and requests type gratuitous arp will trigger the
1081 ARP table to be updated, if this setting is on.
1082
1083 If the ARP table already contains the IP address of the
1084 gratuitous arp frame, the arp table will be updated regardless
1085 if this setting is on or off.
1086
Neil Hormanc1b1bce2006-03-20 22:40:03 -08001087
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088app_solicit - INTEGER
1089 The maximum number of probes to send to the user space ARP daemon
1090 via netlink before dropping back to multicast probes (see
1091 mcast_solicit). Defaults to 0.
1092
1093disable_policy - BOOLEAN
1094 Disable IPSEC policy (SPD) for this interface
1095
1096disable_xfrm - BOOLEAN
1097 Disable IPSEC encryption on this interface, whatever the policy
1098
Hannes Frederic Sowafc4eba52013-08-14 01:03:46 +02001099igmpv2_unsolicited_report_interval - INTEGER
1100 The interval in milliseconds in which the next unsolicited
1101 IGMPv1 or IGMPv2 report retransmit will take place.
1102 Default: 10000 (10 seconds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
Hannes Frederic Sowafc4eba52013-08-14 01:03:46 +02001104igmpv3_unsolicited_report_interval - INTEGER
1105 The interval in milliseconds in which the next unsolicited
1106 IGMPv3 report retransmit will take place.
1107 Default: 1000 (1 seconds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
Martin Schwenked922e1c2014-01-28 15:26:42 +11001109promote_secondaries - BOOLEAN
1110 When a primary IP address is removed from this interface
1111 promote a corresponding secondary IP address instead of
1112 removing all the corresponding secondary IP addresses.
1113
1114
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115tag - INTEGER
1116 Allows you to write a number, which can be used as required.
1117 Default value is 0.
1118
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119Alexey Kuznetsov.
1120kuznet@ms2.inr.ac.ru
1121
1122Updated by:
1123Andi Kleen
1124ak@muc.de
1125Nicolas Delon
1126delon.nicolas@wanadoo.fr
1127
1128
1129
1130
1131/proc/sys/net/ipv6/* Variables:
1132
1133IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1134apply to IPv6 [XXX?].
1135
1136bindv6only - BOOLEAN
1137 Default value for IPV6_V6ONLY socket option,
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001138 which restricts use of the IPv6 socket to IPv6 communication
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 only.
1140 TRUE: disable IPv4-mapped address feature
1141 FALSE: enable IPv4-mapped address feature
1142
Geoffrey Thomasd5c073ca2011-08-22 11:28:57 -07001143 Default: FALSE (as specified in RFC3493)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
Florent Fourcot6444f722014-01-17 17:15:05 +01001145flowlabel_consistency - BOOLEAN
1146 Protect the consistency (and unicity) of flow label.
1147 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1148 flow label manager.
1149 TRUE: enabled
1150 FALSE: disabled
1151 Default: TRUE
1152
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001153auto_flowlabels - BOOLEAN
1154 Automatically generate flow labels based based on a flow hash
1155 of the packet. This allows intermediate devices, such as routers,
1156 to idenfify packet flows for mechanisms like Equal Cost Multipath
1157 Routing (see RFC 6438).
1158 TRUE: enabled
1159 FALSE: disabled
1160 Default: false
1161
FX Le Bail509aba32014-01-07 14:57:27 +01001162anycast_src_echo_reply - BOOLEAN
1163 Controls the use of anycast addresses as source addresses for ICMPv6
1164 echo reply
1165 TRUE: enabled
1166 FALSE: disabled
1167 Default: FALSE
1168
Hannes Frederic Sowa2f711932014-09-02 15:49:25 +02001169mld_qrv - INTEGER
1170 Controls the MLD query robustness variable (see RFC3810 9.1).
1171 Default: 2 (as specified by RFC3810 9.1)
1172 Minimum: 1 (as specified by RFC6636 4.5)
1173
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174IPv6 Fragmentation:
1175
1176ip6frag_high_thresh - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001177 Maximum memory used to reassemble IPv6 fragments. When
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1179 the fragment handler will toss packets until ip6frag_low_thresh
1180 is reached.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182ip6frag_low_thresh - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001183 See ip6frag_high_thresh
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
1185ip6frag_time - INTEGER
1186 Time in seconds to keep an IPv6 fragment in memory.
1187
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188conf/default/*:
1189 Change the interface-specific default settings.
1190
1191
1192conf/all/*:
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001193 Change all the interface-specific settings.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
1195 [XXX: Other special features than forwarding?]
1196
1197conf/all/forwarding - BOOLEAN
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001198 Enable global IPv6 forwarding between all interfaces.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001200 IPv4 and IPv6 work differently here; e.g. netfilter must be used
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 to control which interfaces may forward packets and which not.
1202
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001203 This also sets all interfaces' Host/Router setting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 'forwarding' to the specified value. See below for details.
1205
1206 This referred to as global forwarding.
1207
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -07001208proxy_ndp - BOOLEAN
1209 Do proxy ndp.
1210
Loganaden Velvindron219b5f22014-11-04 03:02:49 -08001211fwmark_reflect - BOOLEAN
1212 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1213 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1214 If unset, these packets have a fwmark of zero. If set, they have the
1215 fwmark of the packet they are replying to.
1216 Default: 0
1217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218conf/interface/*:
1219 Change special settings per interface.
1220
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001221 The functional behaviour for certain settings is different
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 depending on whether local forwarding is enabled or not.
1223
Roy.Li605b91c2011-09-28 19:51:54 +00001224accept_ra - INTEGER
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 Accept Router Advertisements; autoconfigure using them.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001226
Tore Anderson026359b2011-08-28 23:47:33 +00001227 It also determines whether or not to transmit Router
1228 Solicitations. If and only if the functional setting is to
1229 accept Router Advertisements, Router Solicitations will be
1230 transmitted.
1231
Thomas Grafae8abfa2010-09-03 05:47:30 +00001232 Possible values are:
1233 0 Do not accept Router Advertisements.
1234 1 Accept Router Advertisements if forwarding is disabled.
1235 2 Overrule forwarding behaviour. Accept Router Advertisements
1236 even if forwarding is enabled.
1237
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 Functional default: enabled if local forwarding is disabled.
1239 disabled if local forwarding is enabled.
1240
YOSHIFUJI Hideaki65f5c7c2006-03-20 16:55:08 -08001241accept_ra_defrtr - BOOLEAN
1242 Learn default router in Router Advertisement.
1243
1244 Functional default: enabled if accept_ra is enabled.
1245 disabled if accept_ra is disabled.
1246
Ben Greeard9333192014-06-25 14:44:53 -07001247accept_ra_from_local - BOOLEAN
1248 Accept RA with source-address that is found on local machine
1249 if the RA is otherwise proper and able to be accepted.
1250 Default is to NOT accept these as it may be an un-intended
1251 network loop.
1252
1253 Functional default:
1254 enabled if accept_ra_from_local is enabled
1255 on a specific interface.
1256 disabled if accept_ra_from_local is disabled
1257 on a specific interface.
1258
YOSHIFUJI Hideakic4fd30e2006-03-20 16:55:26 -08001259accept_ra_pinfo - BOOLEAN
Matt LaPlante2fe0ae72006-10-03 22:50:39 +02001260 Learn Prefix Information in Router Advertisement.
YOSHIFUJI Hideakic4fd30e2006-03-20 16:55:26 -08001261
1262 Functional default: enabled if accept_ra is enabled.
1263 disabled if accept_ra is disabled.
1264
YOSHIFUJI Hideaki09c884d2006-03-20 17:07:03 -08001265accept_ra_rt_info_max_plen - INTEGER
1266 Maximum prefix length of Route Information in RA.
1267
1268 Route Information w/ prefix larger than or equal to this
1269 variable shall be ignored.
1270
1271 Functional default: 0 if accept_ra_rtr_pref is enabled.
1272 -1 if accept_ra_rtr_pref is disabled.
1273
YOSHIFUJI Hideaki930d6ff2006-03-20 17:05:30 -08001274accept_ra_rtr_pref - BOOLEAN
1275 Accept Router Preference in RA.
1276
1277 Functional default: enabled if accept_ra is enabled.
1278 disabled if accept_ra is disabled.
1279
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280accept_redirects - BOOLEAN
1281 Accept Redirects.
1282
1283 Functional default: enabled if local forwarding is disabled.
1284 disabled if local forwarding is enabled.
1285
YOSHIFUJI Hideaki0bcbc922007-04-24 14:58:30 -07001286accept_source_route - INTEGER
1287 Accept source routing (routing extension header).
1288
YOSHIFUJI Hideakibb4dbf92007-07-10 22:55:49 -07001289 >= 0: Accept only routing header type 2.
YOSHIFUJI Hideaki0bcbc922007-04-24 14:58:30 -07001290 < 0: Do not accept routing header.
1291
1292 Default: 0
1293
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294autoconf - BOOLEAN
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001295 Autoconfigure addresses using Prefix Information in Router
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 Advertisements.
1297
YOSHIFUJI Hideakic4fd30e2006-03-20 16:55:26 -08001298 Functional default: enabled if accept_ra_pinfo is enabled.
1299 disabled if accept_ra_pinfo is disabled.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
1301dad_transmits - INTEGER
1302 The amount of Duplicate Address Detection probes to send.
1303 Default: 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304
Roy.Li605b91c2011-09-28 19:51:54 +00001305forwarding - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001306 Configure interface-specific Host/Router behaviour.
1307
1308 Note: It is recommended to have the same setting on all
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 interfaces; mixed router/host scenarios are rather uncommon.
1310
Thomas Grafae8abfa2010-09-03 05:47:30 +00001311 Possible values are:
1312 0 Forwarding disabled
1313 1 Forwarding enabled
Thomas Grafae8abfa2010-09-03 05:47:30 +00001314
1315 FALSE (0):
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316
1317 By default, Host behaviour is assumed. This means:
1318
1319 1. IsRouter flag is not set in Neighbour Advertisements.
Tore Anderson026359b2011-08-28 23:47:33 +00001320 2. If accept_ra is TRUE (default), transmit Router
1321 Solicitations.
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001322 3. If accept_ra is TRUE (default), accept Router
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 Advertisements (and do autoconfiguration).
1324 4. If accept_redirects is TRUE (default), accept Redirects.
1325
Thomas Grafae8abfa2010-09-03 05:47:30 +00001326 TRUE (1):
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001328 If local forwarding is enabled, Router behaviour is assumed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 This means exactly the reverse from the above:
1330
1331 1. IsRouter flag is set in Neighbour Advertisements.
Tore Anderson026359b2011-08-28 23:47:33 +00001332 2. Router Solicitations are not sent unless accept_ra is 2.
Thomas Grafae8abfa2010-09-03 05:47:30 +00001333 3. Router Advertisements are ignored unless accept_ra is 2.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 4. Redirects are ignored.
1335
Thomas Grafae8abfa2010-09-03 05:47:30 +00001336 Default: 0 (disabled) if global forwarding is disabled (default),
1337 otherwise 1 (enabled).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
1339hop_limit - INTEGER
1340 Default Hop Limit to set.
1341 Default: 64
1342
1343mtu - INTEGER
1344 Default Maximum Transfer Unit
1345 Default: 1280 (IPv6 required minimum)
1346
YOSHIFUJI Hideaki52e16352006-03-20 17:05:47 -08001347router_probe_interval - INTEGER
1348 Minimum interval (in seconds) between Router Probing described
1349 in RFC4191.
1350
1351 Default: 60
1352
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353router_solicitation_delay - INTEGER
1354 Number of seconds to wait after interface is brought up
1355 before sending Router Solicitations.
1356 Default: 1
1357
1358router_solicitation_interval - INTEGER
1359 Number of seconds to wait between Router Solicitations.
1360 Default: 4
1361
1362router_solicitations - INTEGER
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001363 Number of Router Solicitations to send until assuming no
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 routers are present.
1365 Default: 3
1366
1367use_tempaddr - INTEGER
1368 Preference for Privacy Extensions (RFC3041).
1369 <= 0 : disable Privacy Extensions
1370 == 1 : enable Privacy Extensions, but prefer public
1371 addresses over temporary addresses.
1372 > 1 : enable Privacy Extensions and prefer temporary
1373 addresses over public addresses.
1374 Default: 0 (for most devices)
1375 -1 (for point-to-point devices and loopback devices)
1376
1377temp_valid_lft - INTEGER
1378 valid lifetime (in seconds) for temporary addresses.
1379 Default: 604800 (7 days)
1380
1381temp_prefered_lft - INTEGER
1382 Preferred lifetime (in seconds) for temporary addresses.
1383 Default: 86400 (1 day)
1384
1385max_desync_factor - INTEGER
1386 Maximum value for DESYNC_FACTOR, which is a random value
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001387 that ensures that clients don't synchronize with each
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 other and generate new addresses at exactly the same time.
1389 value is in seconds.
1390 Default: 600
Jesper Dangaard Brouere18f5fe2009-02-23 04:39:04 +00001391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392regen_max_retry - INTEGER
1393 Number of attempts before give up attempting to generate
1394 valid temporary addresses.
1395 Default: 5
1396
1397max_addresses - INTEGER
Brian Haleye79dc482010-02-22 12:27:21 +00001398 Maximum number of autoconfigured addresses per interface. Setting
1399 to zero disables the limitation. It is not recommended to set this
1400 value too large (or to zero) because it would be an easy way to
1401 crash the kernel by allowing too many addresses to be created.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 Default: 16
1403
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +09001404disable_ipv6 - BOOLEAN
Brian Haley9bdd8d42009-03-18 18:22:48 -07001405 Disable IPv6 operation. If accept_dad is set to 2, this value
1406 will be dynamically set to TRUE if DAD fails for the link-local
1407 address.
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +09001408 Default: FALSE (enable IPv6 operation)
1409
Brian Haley56d417b2009-06-01 03:07:33 -07001410 When this value is changed from 1 to 0 (IPv6 is being enabled),
1411 it will dynamically create a link-local address on the given
1412 interface and start Duplicate Address Detection, if necessary.
1413
1414 When this value is changed from 0 to 1 (IPv6 is being disabled),
1415 it will dynamically delete all address on the given interface.
1416
YOSHIFUJI Hideaki1b34be72008-06-28 14:18:38 +09001417accept_dad - INTEGER
1418 Whether to accept DAD (Duplicate Address Detection).
1419 0: Disable DAD
1420 1: Enable DAD (default)
1421 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1422 link-local address has been found.
1423
Octavian Purdilaf7734fd2009-10-02 11:39:15 +00001424force_tllao - BOOLEAN
1425 Enable sending the target link-layer address option even when
1426 responding to a unicast neighbor solicitation.
1427 Default: FALSE
1428
1429 Quoting from RFC 2461, section 4.4, Target link-layer address:
1430
1431 "The option MUST be included for multicast solicitations in order to
1432 avoid infinite Neighbor Solicitation "recursion" when the peer node
1433 does not have a cache entry to return a Neighbor Advertisements
1434 message. When responding to unicast solicitations, the option can be
1435 omitted since the sender of the solicitation has the correct link-
1436 layer address; otherwise it would not have be able to send the unicast
1437 solicitation in the first place. However, including the link-layer
1438 address in this case adds little overhead and eliminates a potential
1439 race condition where the sender deletes the cached link-layer address
1440 prior to receiving a response to a previous solicitation."
1441
Hannes Frederic Sowadb2b6202013-01-01 00:35:31 +00001442ndisc_notify - BOOLEAN
1443 Define mode for notification of address and device changes.
1444 0 - (default): do nothing
1445 1 - Generate unsolicited neighbour advertisements when device is brought
1446 up or hardware address changes.
1447
Hannes Frederic Sowafc4eba52013-08-14 01:03:46 +02001448mldv1_unsolicited_report_interval - INTEGER
1449 The interval in milliseconds in which the next unsolicited
1450 MLDv1 report retransmit will take place.
1451 Default: 10000 (10 seconds)
1452
1453mldv2_unsolicited_report_interval - INTEGER
1454 The interval in milliseconds in which the next unsolicited
1455 MLDv2 report retransmit will take place.
1456 Default: 1000 (1 second)
1457
Daniel Borkmannf2127812013-09-04 00:19:44 +02001458force_mld_version - INTEGER
1459 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1460 1 - Enforce to use MLD version 1
1461 2 - Enforce to use MLD version 2
1462
Hannes Frederic Sowab800c3b2013-08-27 01:36:51 +02001463suppress_frag_ndisc - INTEGER
1464 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1465 with IPv6 Neighbor Discovery) behavior:
1466 1 - (default) discard fragmented neighbor discovery packets
1467 0 - allow fragmented neighbor discovery packets
1468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469icmp/*:
1470ratelimit - INTEGER
1471 Limit the maximal rates for sending ICMPv6 packets.
Stephen Hemminger6dbf4bc2008-07-01 19:29:07 -07001472 0 to disable any limiting,
1473 otherwise the minimal space between responses in milliseconds.
1474 Default: 1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475
1476
1477IPv6 Update by:
1478Pekka Savola <pekkas@netcore.fi>
1479YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1480
1481
1482/proc/sys/net/bridge/* Variables:
1483
1484bridge-nf-call-arptables - BOOLEAN
1485 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1486 0 : disable this.
1487 Default: 1
1488
1489bridge-nf-call-iptables - BOOLEAN
1490 1 : pass bridged IPv4 traffic to iptables' chains.
1491 0 : disable this.
1492 Default: 1
1493
1494bridge-nf-call-ip6tables - BOOLEAN
1495 1 : pass bridged IPv6 traffic to ip6tables' chains.
1496 0 : disable this.
1497 Default: 1
1498
1499bridge-nf-filter-vlan-tagged - BOOLEAN
Michael Milner516299d2007-04-12 22:14:23 -07001500 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1501 0 : disable this.
Pablo Neira Ayuso49816822012-05-08 19:36:44 +02001502 Default: 0
Michael Milner516299d2007-04-12 22:14:23 -07001503
1504bridge-nf-filter-pppoe-tagged - BOOLEAN
1505 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 0 : disable this.
Pablo Neira Ayuso49816822012-05-08 19:36:44 +02001507 Default: 0
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
Pablo Neira Ayuso49816822012-05-08 19:36:44 +02001509bridge-nf-pass-vlan-input-dev - BOOLEAN
1510 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1511 interface on the bridge and set the netfilter input device to the vlan.
1512 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1513 target work with vlan-on-top-of-bridge interfaces. When no matching
1514 vlan interface is found, or this switch is off, the input device is
1515 set to the bridge interface.
1516 0: disable bridge netfilter vlan interface lookup.
1517 Default: 0
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001519proc/sys/net/sctp/* Variables:
1520
1521addip_enable - BOOLEAN
1522 Enable or disable extension of Dynamic Address Reconfiguration
1523 (ADD-IP) functionality specified in RFC5061. This extension provides
1524 the ability to dynamically add and remove new addresses for the SCTP
1525 associations.
1526
1527 1: Enable extension.
1528
1529 0: Disable extension.
1530
1531 Default: 0
1532
1533addip_noauth_enable - BOOLEAN
1534 Dynamic Address Reconfiguration (ADD-IP) requires the use of
1535 authentication to protect the operations of adding or removing new
1536 addresses. This requirement is mandated so that unauthorized hosts
1537 would not be able to hijack associations. However, older
1538 implementations may not have implemented this requirement while
1539 allowing the ADD-IP extension. For reasons of interoperability,
1540 we provide this variable to control the enforcement of the
1541 authentication requirement.
1542
1543 1: Allow ADD-IP extension to be used without authentication. This
1544 should only be set in a closed environment for interoperability
1545 with older implementations.
1546
1547 0: Enforce the authentication requirement
1548
1549 Default: 0
1550
1551auth_enable - BOOLEAN
1552 Enable or disable Authenticated Chunks extension. This extension
1553 provides the ability to send and receive authenticated chunks and is
1554 required for secure operation of Dynamic Address Reconfiguration
1555 (ADD-IP) extension.
1556
1557 1: Enable this extension.
1558 0: Disable this extension.
1559
1560 Default: 0
1561
1562prsctp_enable - BOOLEAN
1563 Enable or disable the Partial Reliability extension (RFC3758) which
1564 is used to notify peers that a given DATA should no longer be expected.
1565
1566 1: Enable extension
1567 0: Disable
1568
1569 Default: 1
1570
1571max_burst - INTEGER
1572 The limit of the number of new packets that can be initially sent. It
1573 controls how bursty the generated traffic can be.
1574
1575 Default: 4
1576
1577association_max_retrans - INTEGER
1578 Set the maximum number for retransmissions that an association can
1579 attempt deciding that the remote end is unreachable. If this value
1580 is exceeded, the association is terminated.
1581
1582 Default: 10
1583
1584max_init_retransmits - INTEGER
1585 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1586 that an association will attempt before declaring the destination
1587 unreachable and terminating.
1588
1589 Default: 8
1590
1591path_max_retrans - INTEGER
1592 The maximum number of retransmissions that will be attempted on a given
1593 path. Once this threshold is exceeded, the path is considered
1594 unreachable, and new traffic will use a different path when the
1595 association is multihomed.
1596
1597 Default: 5
1598
Neil Horman5aa93bc2012-07-21 07:56:07 +00001599pf_retrans - INTEGER
1600 The number of retransmissions that will be attempted on a given path
1601 before traffic is redirected to an alternate transport (should one
1602 exist). Note this is distinct from path_max_retrans, as a path that
1603 passes the pf_retrans threshold can still be used. Its only
1604 deprioritized when a transmission path is selected by the stack. This
1605 setting is primarily used to enable fast failover mechanisms without
1606 having to reduce path_max_retrans to a very low value. See:
1607 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1608 for details. Note also that a value of pf_retrans > path_max_retrans
1609 disables this feature
1610
1611 Default: 0
1612
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001613rto_initial - INTEGER
1614 The initial round trip timeout value in milliseconds that will be used
1615 in calculating round trip times. This is the initial time interval
1616 for retransmissions.
1617
1618 Default: 3000
1619
1620rto_max - INTEGER
1621 The maximum value (in milliseconds) of the round trip timeout. This
1622 is the largest time interval that can elapse between retransmissions.
1623
1624 Default: 60000
1625
1626rto_min - INTEGER
1627 The minimum value (in milliseconds) of the round trip timeout. This
1628 is the smallest time interval the can elapse between retransmissions.
1629
1630 Default: 1000
1631
1632hb_interval - INTEGER
1633 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
1634 are sent at the specified interval on idle paths to probe the state of
1635 a given path between 2 associations.
1636
1637 Default: 30000
1638
1639sack_timeout - INTEGER
1640 The amount of time (in milliseconds) that the implementation will wait
1641 to send a SACK.
1642
1643 Default: 200
1644
1645valid_cookie_life - INTEGER
1646 The default lifetime of the SCTP cookie (in milliseconds). The cookie
1647 is used during association establishment.
1648
1649 Default: 60000
1650
1651cookie_preserve_enable - BOOLEAN
1652 Enable or disable the ability to extend the lifetime of the SCTP cookie
1653 that is used during the establishment phase of SCTP association
1654
1655 1: Enable cookie lifetime extension.
1656 0: Disable
1657
1658 Default: 1
1659
Neil Horman3c681982012-10-24 09:20:03 +00001660cookie_hmac_alg - STRING
1661 Select the hmac algorithm used when generating the cookie value sent by
1662 a listening sctp socket to a connecting client in the INIT-ACK chunk.
1663 Valid values are:
1664 * md5
1665 * sha1
1666 * none
1667 Ability to assign md5 or sha1 as the selected alg is predicated on the
stephen hemminger3b09adc2013-01-03 07:50:29 +00001668 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
Neil Horman3c681982012-10-24 09:20:03 +00001669 CONFIG_CRYPTO_SHA1).
1670
1671 Default: Dependent on configuration. MD5 if available, else SHA1 if
1672 available, else none.
1673
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001674rcvbuf_policy - INTEGER
1675 Determines if the receive buffer is attributed to the socket or to
1676 association. SCTP supports the capability to create multiple
1677 associations on a single socket. When using this capability, it is
1678 possible that a single stalled association that's buffering a lot
1679 of data may block other associations from delivering their data by
1680 consuming all of the receive buffer space. To work around this,
1681 the rcvbuf_policy could be set to attribute the receiver buffer space
1682 to each association instead of the socket. This prevents the described
1683 blocking.
1684
1685 1: rcvbuf space is per association
stephen hemminger3b09adc2013-01-03 07:50:29 +00001686 0: rcvbuf space is per socket
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001687
1688 Default: 0
1689
1690sndbuf_policy - INTEGER
1691 Similar to rcvbuf_policy above, this applies to send buffer space.
1692
1693 1: Send buffer is tracked per association
1694 0: Send buffer is tracked per socket.
1695
1696 Default: 0
1697
1698sctp_mem - vector of 3 INTEGERs: min, pressure, max
1699 Number of pages allowed for queueing by all SCTP sockets.
1700
1701 min: Below this number of pages SCTP is not bothered about its
1702 memory appetite. When amount of memory allocated by SCTP exceeds
1703 this number, SCTP starts to moderate memory usage.
1704
1705 pressure: This value was introduced to follow format of tcp_mem.
1706
1707 max: Number of pages allowed for queueing by all SCTP sockets.
1708
1709 Default is calculated at boot time from amount of available memory.
1710
1711sctp_rmem - vector of 3 INTEGERs: min, default, max
Max Matveeva6e12042011-06-19 22:08:10 +00001712 Only the first value ("min") is used, "default" and "max" are
1713 ignored.
1714
1715 min: Minimal size of receive buffer used by SCTP socket.
1716 It is guaranteed to each SCTP socket (but not association) even
1717 under moderate memory pressure.
1718
1719 Default: 1 page
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001720
1721sctp_wmem - vector of 3 INTEGERs: min, default, max
Max Matveeva6e12042011-06-19 22:08:10 +00001722 Currently this tunable has no effect.
Vlad Yasevich32e8d492008-07-08 16:43:29 -07001723
Bhaskar Dutta72388432009-09-03 17:25:47 +05301724addr_scope_policy - INTEGER
1725 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1726
1727 0 - Disable IPv4 address scoping
1728 1 - Enable IPv4 address scoping
1729 2 - Follow draft but allow IPv4 private addresses
1730 3 - Follow draft but allow IPv4 link local addresses
1731
1732 Default: 1
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734
Stephen Hemminger4edc2f32008-07-10 16:50:26 -07001735/proc/sys/net/core/*
Shan Weic60f6aa2012-04-26 16:52:52 +00001736 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
Wang Tinggong705efc32009-05-14 22:49:36 +00001737
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
Stephen Hemminger4edc2f32008-07-10 16:50:26 -07001739/proc/sys/net/unix/*
Wang Tinggong705efc32009-05-14 22:49:36 +00001740max_dgram_qlen - INTEGER
1741 The maximum length of dgram socket receive queue
1742
1743 Default: 10
1744
1745
1746UNDOCUMENTED:
Stephen Hemminger4edc2f32008-07-10 16:50:26 -07001747
1748/proc/sys/net/irda/*
1749 fast_poll_increase FIXME
1750 warn_noreply_time FIXME
1751 discovery_slots FIXME
1752 slot_timeout FIXME
1753 max_baud_rate FIXME
1754 discovery_timeout FIXME
1755 lap_keepalive_time FIXME
1756 max_noreply_time FIXME
1757 max_tx_data_size FIXME
1758 max_tx_window FIXME
1759 min_tx_turn_time FIXME