1 /proc/sys/net/ipv4/* Variables:
7 Forward Packets between interfaces.
9 This variable is special, its change resets all configuration
10 parameters to their default state (RFC1122 for hosts, RFC1812
13 ip_default_ttl - INTEGER
14 Default value of TTL field (Time To Live) for outgoing (but not
15 forwarded) IP packets. Should be between 1 and 255 inclusive.
16 Default: 64 (as recommended by RFC1700)
18 ip_no_pmtu_disc - INTEGER
19 Disable Path MTU Discovery. If enabled in mode 1 and a
20 fragmentation-required ICMP is received, the PMTU to this
21 destination will be set to min_pmtu (see below). You will need
22 to raise min_pmtu to the smallest interface MTU on your system
23 manually if you want to avoid locally generated fragments.
25 In mode 2 incoming Path MTU Discovery messages will be
26 discarded. Outgoing frames are handled the same as in mode 1,
27 implicitly setting IP_PMTUDISC_DONT on every created socket.
29 Mode 3 is a hardened pmtu discover mode. The kernel will only
30 accept fragmentation-needed errors if the underlying protocol
31 can verify them besides a plain socket lookup. Current
32 protocols for which pmtu events will be honored are TCP, SCTP
33 and DCCP as they verify e.g. the sequence number or the
34 association. This mode should not be enabled globally but is
35 only intended to secure e.g. name servers in namespaces where
36 TCP path mtu must still work but path MTU information of other
37 protocols should be discarded. If enabled globally this mode
38 could break other protocols.
44 default 552 - minimum discovered Path MTU
46 ip_forward_use_pmtu - BOOLEAN
47 By default we don't trust protocol path MTUs while forwarding
48 because they could be easily forged and can lead to unwanted
49 fragmentation by the router.
50 You only need to enable this if you have user-space software
51 which tries to discover path mtus by itself and depends on the
52 kernel honoring this information. This is normally not the
59 fwmark_reflect - BOOLEAN
60 Controls the fwmark of kernel-generated IPv4 reply packets that are not
61 associated with a socket for example, TCP RSTs or ICMP echo replies).
62 If unset, these packets have a fwmark of zero. If set, they have the
63 fwmark of the packet they are replying to.
66 fib_multipath_use_neigh - BOOLEAN
67 Use status of existing neighbor entry when determining nexthop for
68 multipath routes. If disabled, neighbor information is not used and
69 packets could be directed to a failed nexthop. Only valid for kernels
70 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
76 fib_multipath_hash_policy - INTEGER
77 Controls which hash policy to use for multipath routes. Only valid
78 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
84 ip_forward_update_priority - INTEGER
85 Whether to update SKB priority from "TOS" field in IPv4 header after it
86 is forwarded. The new SKB priority is mapped from TOS field value
87 according to an rt_tos2priority table (see e.g. man tc-prio).
88 Default: 1 (Update priority.)
90 0 - Do not update priority.
93 route/max_size - INTEGER
94 Maximum number of routes allowed in the kernel. Increase
95 this when using large numbers of interfaces and/or routes.
96 From linux kernel 3.6 onwards, this is deprecated for ipv4
97 as route cache is no longer used.
99 neigh/default/gc_thresh1 - INTEGER
100 Minimum number of entries to keep. Garbage collector will not
101 purge entries if there are fewer than this number.
104 neigh/default/gc_thresh2 - INTEGER
105 Threshold when garbage collector becomes more aggressive about
106 purging entries. Entries older than 5 seconds will be cleared
107 when over this number.
110 neigh/default/gc_thresh3 - INTEGER
111 Maximum number of non-PERMANENT neighbor entries allowed. Increase
112 this when using large numbers of interfaces and when communicating
113 with large numbers of directly-connected peers.
116 neigh/default/unres_qlen_bytes - INTEGER
117 The maximum number of bytes which may be used by packets
118 queued for each unresolved address by other network layers.
120 Setting negative value is meaningless and will return error.
121 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
122 Exact value depends on architecture and kernel options,
123 but should be enough to allow queuing 256 packets
126 neigh/default/unres_qlen - INTEGER
127 The maximum number of packets which may be queued for each
128 unresolved address by other network layers.
129 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
130 Prior to linux 3.3, the default value is 3 which may cause
131 unexpected packet loss. The current default value is calculated
132 according to default value of unres_qlen_bytes and true size of
136 mtu_expires - INTEGER
137 Time, in seconds, that cached PMTU information is kept.
139 min_adv_mss - INTEGER
140 The advertised MSS depends on the first hop route MTU, but will
141 never be lower than this setting.
145 ipfrag_high_thresh - LONG INTEGER
146 Maximum memory used to reassemble IP fragments.
148 ipfrag_low_thresh - LONG INTEGER
149 (Obsolete since linux-4.17)
150 Maximum memory used to reassemble IP fragments before the kernel
151 begins to remove incomplete fragment queues to free up resources.
152 The kernel still accepts new fragments for defragmentation.
154 ipfrag_time - INTEGER
155 Time in seconds to keep an IP fragment in memory.
157 ipfrag_max_dist - INTEGER
158 ipfrag_max_dist is a non-negative integer value which defines the
159 maximum "disorder" which is allowed among fragments which share a
160 common IP source address. Note that reordering of packets is
161 not unusual, but if a large number of fragments arrive from a source
162 IP address while a particular fragment queue remains incomplete, it
163 probably indicates that one or more fragments belonging to that queue
164 have been lost. When ipfrag_max_dist is positive, an additional check
165 is done on fragments before they are added to a reassembly queue - if
166 ipfrag_max_dist (or more) fragments have arrived from a particular IP
167 address between additions to any IP fragment queue using that source
168 address, it's presumed that one or more fragments in the queue are
169 lost. The existing fragment queue will be dropped, and a new one
170 started. An ipfrag_max_dist value of zero disables this check.
172 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
173 result in unnecessarily dropping fragment queues when normal
174 reordering of packets occurs, which could lead to poor application
175 performance. Using a very large value, e.g. 50000, increases the
176 likelihood of incorrectly reassembling IP fragments that originate
177 from different IP datagrams, which could result in data corruption.
182 inet_peer_threshold - INTEGER
183 The approximate size of the storage. Starting from this threshold
184 entries will be thrown aggressively. This threshold also determines
185 entries' time-to-live and time intervals between garbage collection
186 passes. More entries, less time-to-live, less GC interval.
188 inet_peer_minttl - INTEGER
189 Minimum time-to-live of entries. Should be enough to cover fragment
190 time-to-live on the reassembling side. This minimum time-to-live is
191 guaranteed if the pool size is less than inet_peer_threshold.
194 inet_peer_maxttl - INTEGER
195 Maximum time-to-live of entries. Unused entries will expire after
196 this period of time if there is no memory pressure on the pool (i.e.
197 when the number of entries in the pool is very small).
203 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
204 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
207 tcp_abort_on_overflow - BOOLEAN
208 If listening service is too slow to accept new connections,
209 reset them. Default state is FALSE. It means that if overflow
210 occurred due to a burst, connection will recover. Enable this
211 option _only_ if you are really sure that listening daemon
212 cannot be tuned to accept connections faster. Enabling this
213 option can harm clients of your server.
215 tcp_adv_win_scale - INTEGER
216 Count buffering overhead as bytes/2^tcp_adv_win_scale
217 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
219 Possible values are [-31, 31], inclusive.
222 tcp_allowed_congestion_control - STRING
223 Show/set the congestion control choices available to non-privileged
224 processes. The list is a subset of those listed in
225 tcp_available_congestion_control.
226 Default is "reno" and the default setting (tcp_congestion_control).
228 tcp_app_win - INTEGER
229 Reserve max(window/2^tcp_app_win, mss) of window for application
230 buffer. Value 0 is special, it means that nothing is reserved.
233 tcp_autocorking - BOOLEAN
234 Enable TCP auto corking :
235 When applications do consecutive small write()/sendmsg() system calls,
236 we try to coalesce these small writes as much as possible, to lower
237 total amount of sent packets. This is done if at least one prior
238 packet for the flow is waiting in Qdisc queues or device transmit
239 queue. Applications can still use TCP_CORK for optimal behavior
240 when they know how/when to uncork their sockets.
243 tcp_available_congestion_control - STRING
244 Shows the available congestion control choices that are registered.
245 More congestion control algorithms may be available as modules,
248 tcp_base_mss - INTEGER
249 The initial value of search_low to be used by the packetization layer
250 Path MTU discovery (MTU probing). If MTU probing is enabled,
251 this is the initial MSS used by the connection.
253 tcp_congestion_control - STRING
254 Set the congestion control algorithm to be used for new
255 connections. The algorithm "reno" is always available, but
256 additional choices may be available based on kernel configuration.
257 Default is set as part of kernel configuration.
258 For passive connections, the listener congestion control choice
260 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
263 Allows TCP to send "duplicate" SACKs.
265 tcp_early_retrans - INTEGER
266 Tail loss probe (TLP) converts RTOs occurring due to tail
267 losses into fast recovery (draft-ietf-tcpm-rack). Note that
268 TLP requires RACK to function properly (see tcp_recovery below)
275 Control use of Explicit Congestion Notification (ECN) by TCP.
276 ECN is used only when both ends of the TCP connection indicate
277 support for it. This feature is useful in avoiding losses due
278 to congestion by allowing supporting routers to signal
279 congestion before having to drop packets.
281 0 Disable ECN. Neither initiate nor accept ECN.
282 1 Enable ECN when requested by incoming connections and
283 also request ECN on outgoing connection attempts.
284 2 Enable ECN when requested by incoming connections
285 but do not request ECN on outgoing connections.
288 tcp_ecn_fallback - BOOLEAN
289 If the kernel detects that ECN connection misbehaves, enable fall
290 back to non-ECN. Currently, this knob implements the fallback
291 from RFC3168, section 6.1.1.1., but we reserve that in future,
292 additional detection mechanisms could be implemented under this
293 knob. The value is not used, if tcp_ecn or per route (or congestion
294 control) ECN settings are disabled.
295 Default: 1 (fallback enabled)
298 This is a legacy option, it has no effect anymore.
300 tcp_fin_timeout - INTEGER
301 The length of time an orphaned (no longer referenced by any
302 application) connection will remain in the FIN_WAIT_2 state
303 before it is aborted at the local end. While a perfectly
304 valid "receive only" state for an un-orphaned connection, an
305 orphaned connection in FIN_WAIT_2 state could otherwise wait
306 forever for the remote to close its end of the connection.
311 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
312 F-RTO is an enhanced recovery algorithm for TCP retransmission
313 timeouts. It is particularly beneficial in networks where the
314 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
315 modification. It does not require any support from the peer.
317 By default it's enabled with a non-zero value. 0 disables F-RTO.
319 tcp_fwmark_accept - BOOLEAN
320 If set, incoming connections to listening sockets that do not have a
321 socket mark will set the mark of the accepting socket to the fwmark of
322 the incoming SYN packet. This will cause all packets on that connection
323 (starting from the first SYNACK) to be sent with that fwmark. The
324 listening socket's mark is unchanged. Listening sockets that already
325 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
330 tcp_invalid_ratelimit - INTEGER
331 Limit the maximal rate for sending duplicate acknowledgments
332 in response to incoming TCP packets that are for an existing
333 connection but that are invalid due to any of these reasons:
335 (a) out-of-window sequence number,
336 (b) out-of-window acknowledgment number, or
337 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
339 This can help mitigate simple "ack loop" DoS attacks, wherein
340 a buggy or malicious middlebox or man-in-the-middle can
341 rewrite TCP header fields in manner that causes each endpoint
342 to think that the other is sending invalid TCP segments, thus
343 causing each side to send an unterminating stream of duplicate
344 acknowledgments for invalid segments.
346 Using 0 disables rate-limiting of dupacks in response to
347 invalid segments; otherwise this value specifies the minimal
348 space between sending such dupacks, in milliseconds.
350 Default: 500 (milliseconds).
352 tcp_keepalive_time - INTEGER
353 How often TCP sends out keepalive messages when keepalive is enabled.
356 tcp_keepalive_probes - INTEGER
357 How many keepalive probes TCP sends out, until it decides that the
358 connection is broken. Default value: 9.
360 tcp_keepalive_intvl - INTEGER
361 How frequently the probes are send out. Multiplied by
362 tcp_keepalive_probes it is time to kill not responding connection,
363 after probes started. Default value: 75sec i.e. connection
364 will be aborted after ~11 minutes of retries.
366 tcp_l3mdev_accept - BOOLEAN
367 Enables child sockets to inherit the L3 master device index.
368 Enabling this option allows a "global" listen socket to work
369 across L3 master domains (e.g., VRFs) with connected sockets
370 derived from the listen socket to be bound to the L3 domain in
371 which the packets originated. Only valid when the kernel was
372 compiled with CONFIG_NET_L3_MASTER_DEV.
373 Default: 0 (disabled)
375 tcp_low_latency - BOOLEAN
376 This is a legacy option, it has no effect anymore.
378 tcp_max_orphans - INTEGER
379 Maximal number of TCP sockets not attached to any user file handle,
380 held by system. If this number is exceeded orphaned connections are
381 reset immediately and warning is printed. This limit exists
382 only to prevent simple DoS attacks, you _must_ not rely on this
383 or lower the limit artificially, but rather increase it
384 (probably, after increasing installed memory),
385 if network conditions require more than default value,
386 and tune network services to linger and kill such states
387 more aggressively. Let me to remind again: each orphan eats
388 up to ~64K of unswappable memory.
390 tcp_max_syn_backlog - INTEGER
391 Maximal number of remembered connection requests, which have not
392 received an acknowledgment from connecting client.
393 The minimal value is 128 for low memory machines, and it will
394 increase in proportion to the memory of machine.
395 If server suffers from overload, try increasing this number.
397 tcp_max_tw_buckets - INTEGER
398 Maximal number of timewait sockets held by system simultaneously.
399 If this number is exceeded time-wait socket is immediately destroyed
400 and warning is printed. This limit exists only to prevent
401 simple DoS attacks, you _must_ not lower the limit artificially,
402 but rather increase it (probably, after increasing installed memory),
403 if network conditions require more than default value.
405 tcp_mem - vector of 3 INTEGERs: min, pressure, max
406 min: below this number of pages TCP is not bothered about its
409 pressure: when amount of memory allocated by TCP exceeds this number
410 of pages, TCP moderates its memory consumption and enters memory
411 pressure mode, which is exited when memory consumption falls
414 max: number of pages allowed for queueing by all TCP sockets.
416 Defaults are calculated at boot time from amount of available
419 tcp_min_rtt_wlen - INTEGER
420 The window length of the windowed min filter to track the minimum RTT.
421 A shorter window lets a flow more quickly pick up new (higher)
422 minimum RTT when it is moved to a longer path (e.g., due to traffic
423 engineering). A longer window makes the filter more resistant to RTT
424 inflations such as transient congestion. The unit is seconds.
425 Possible values: 0 - 86400 (1 day)
428 tcp_moderate_rcvbuf - BOOLEAN
429 If set, TCP performs receive buffer auto-tuning, attempting to
430 automatically size the buffer (no greater than tcp_rmem[2]) to
431 match the size required by the path for full throughput. Enabled by
434 tcp_mtu_probing - INTEGER
435 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
438 1 - Disabled by default, enabled when an ICMP black hole detected
439 2 - Always enabled, use initial MSS of tcp_base_mss.
441 tcp_probe_interval - UNSIGNED INTEGER
442 Controls how often to start TCP Packetization-Layer Path MTU
443 Discovery reprobe. The default is reprobing every 10 minutes as
446 tcp_probe_threshold - INTEGER
447 Controls when TCP Packetization-Layer Path MTU Discovery probing
448 will stop in respect to the width of search range in bytes. Default
451 tcp_no_metrics_save - BOOLEAN
452 By default, TCP saves various connection metrics in the route cache
453 when the connection closes, so that connections established in the
454 near future can use these to set initial conditions. Usually, this
455 increases overall performance, but may sometimes cause performance
456 degradation. If set, TCP will not cache metrics on closing
459 tcp_orphan_retries - INTEGER
460 This value influences the timeout of a locally closed TCP connection,
461 when RTO retransmissions remain unacknowledged.
462 See tcp_retries2 for more details.
464 The default value is 8.
465 If your machine is a loaded WEB server,
466 you should think about lowering this value, such sockets
467 may consume significant resources. Cf. tcp_max_orphans.
469 tcp_recovery - INTEGER
470 This value is a bitmap to enable various experimental loss recovery
473 RACK: 0x1 enables the RACK loss detection for fast detection of lost
474 retransmissions and tail drops. It also subsumes and disables
475 RFC6675 recovery for SACK connections.
476 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
477 RACK: 0x4 disables RACK's DUPACK threshold heuristic
481 tcp_reordering - INTEGER
482 Initial reordering level of packets in a TCP stream.
483 TCP stack can then dynamically adjust flow reordering level
484 between this initial value and tcp_max_reordering
487 tcp_max_reordering - INTEGER
488 Maximal reordering level of packets in a TCP stream.
489 300 is a fairly conservative value, but you might increase it
490 if paths are using per packet load balancing (like bonding rr mode)
493 tcp_retrans_collapse - BOOLEAN
494 Bug-to-bug compatibility with some broken printers.
495 On retransmit try to send bigger packets to work around bugs in
498 tcp_retries1 - INTEGER
499 This value influences the time, after which TCP decides, that
500 something is wrong due to unacknowledged RTO retransmissions,
501 and reports this suspicion to the network layer.
502 See tcp_retries2 for more details.
504 RFC 1122 recommends at least 3 retransmissions, which is the
507 tcp_retries2 - INTEGER
508 This value influences the timeout of an alive TCP connection,
509 when RTO retransmissions remain unacknowledged.
510 Given a value of N, a hypothetical TCP connection following
511 exponential backoff with an initial RTO of TCP_RTO_MIN would
512 retransmit N times before killing the connection at the (N+1)th RTO.
514 The default value of 15 yields a hypothetical timeout of 924.6
515 seconds and is a lower bound for the effective timeout.
516 TCP will effectively time out at the first RTO which exceeds the
517 hypothetical timeout.
519 RFC 1122 recommends at least 100 seconds for the timeout,
520 which corresponds to a value of at least 8.
522 tcp_rfc1337 - BOOLEAN
523 If set, the TCP stack behaves conforming to RFC1337. If unset,
524 we are not conforming to RFC, but prevent TCP TIME_WAIT
528 tcp_rmem - vector of 3 INTEGERs: min, default, max
529 min: Minimal size of receive buffer used by TCP sockets.
530 It is guaranteed to each TCP socket, even under moderate memory
534 default: initial size of receive buffer used by TCP sockets.
535 This value overrides net.core.rmem_default used by other protocols.
536 Default: 87380 bytes. This value results in window of 65535 with
537 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
538 less for default tcp_app_win. See below about these variables.
540 max: maximal size of receive buffer allowed for automatically
541 selected receiver buffers for TCP socket. This value does not override
542 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
543 automatic tuning of that socket's receive buffer size, in which
544 case this value is ignored.
545 Default: between 87380B and 6MB, depending on RAM size.
548 Enable select acknowledgments (SACKS).
550 tcp_comp_sack_delay_ns - LONG INTEGER
551 TCP tries to reduce number of SACK sent, using a timer
552 based on 5% of SRTT, capped by this sysctl, in nano seconds.
553 The default is 1ms, based on TSO autosizing period.
555 Default : 1,000,000 ns (1 ms)
557 tcp_comp_sack_nr - INTEGER
558 Max numer of SACK that can be compressed.
559 Using 0 disables SACK compression.
563 tcp_slow_start_after_idle - BOOLEAN
564 If set, provide RFC2861 behavior and time out the congestion
565 window after an idle period. An idle period is defined at
566 the current RTO. If unset, the congestion window will not
567 be timed out after an idle period.
571 Use the Host requirements interpretation of the TCP urgent pointer field.
572 Most hosts use the older BSD interpretation, so if you turn this on
573 Linux might not communicate correctly with them.
576 tcp_synack_retries - INTEGER
577 Number of times SYNACKs for a passive TCP connection attempt will
578 be retransmitted. Should not be higher than 255. Default value
579 is 5, which corresponds to 31seconds till the last retransmission
580 with the current initial RTO of 1second. With this the final timeout
581 for a passive TCP connection will happen after 63seconds.
583 tcp_syncookies - BOOLEAN
584 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
585 Send out syncookies when the syn backlog queue of a socket
586 overflows. This is to prevent against the common 'SYN flood attack'
589 Note, that syncookies is fallback facility.
590 It MUST NOT be used to help highly loaded servers to stand
591 against legal connection rate. If you see SYN flood warnings
592 in your logs, but investigation shows that they occur
593 because of overload with legal connections, you should tune
594 another parameters until this warning disappear.
595 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
597 syncookies seriously violate TCP protocol, do not allow
598 to use TCP extensions, can result in serious degradation
599 of some services (f.e. SMTP relaying), visible not by you,
600 but your clients and relays, contacting you. While you see
601 SYN flood warnings in logs not being really flooded, your server
602 is seriously misconfigured.
604 If you want to test which effects syncookies have to your
605 network connections you can set this knob to 2 to enable
606 unconditionally generation of syncookies.
608 tcp_fastopen - INTEGER
609 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
612 The client support is enabled by flag 0x1 (on by default). The client
613 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
614 rather than connect() to send data in SYN.
616 The server support is enabled by flag 0x2 (off by default). Then
617 either enable for all listeners with another flag (0x400) or
618 enable individual listeners via TCP_FASTOPEN socket option with
619 the option value being the length of the syn-data backlog.
621 The values (bitmap) are
622 0x1: (client) enables sending data in the opening SYN on the client.
623 0x2: (server) enables the server support, i.e., allowing data in
624 a SYN packet to be accepted and passed to the
625 application before 3-way handshake finishes.
626 0x4: (client) send data in the opening SYN regardless of cookie
627 availability and without a cookie option.
628 0x200: (server) accept data-in-SYN w/o any cookie option present.
629 0x400: (server) enable all listeners to support Fast Open by
630 default without explicit TCP_FASTOPEN socket option.
634 Note that that additional client or server features are only
635 effective if the basic support (0x1 and 0x2) are enabled respectively.
637 tcp_fastopen_blackhole_timeout_sec - INTEGER
638 Initial time period in second to disable Fastopen on active TCP sockets
639 when a TFO firewall blackhole issue happens.
640 This time period will grow exponentially when more blackhole issues
641 get detected right after Fastopen is re-enabled and will reset to
642 initial value when the blackhole issue goes away.
643 0 to disable the blackhole detection.
644 By default, it is set to 1hr.
646 tcp_syn_retries - INTEGER
647 Number of times initial SYNs for an active TCP connection attempt
648 will be retransmitted. Should not be higher than 127. Default value
649 is 6, which corresponds to 63seconds till the last retransmission
650 with the current initial RTO of 1second. With this the final timeout
651 for an active TCP connection attempt will happen after 127seconds.
653 tcp_timestamps - INTEGER
654 Enable timestamps as defined in RFC1323.
656 1: Enable timestamps as defined in RFC1323 and use random offset for
657 each connection rather than only using the current time.
658 2: Like 1, but without random offsets.
661 tcp_min_tso_segs - INTEGER
662 Minimal number of segments per TSO frame.
663 Since linux-3.12, TCP does an automatic sizing of TSO frames,
664 depending on flow rate, instead of filling 64Kbytes packets.
665 For specific usages, it's possible to force TCP to build big
666 TSO frames. Note that TCP stack might split too big TSO packets
667 if available window is too small.
670 tcp_pacing_ss_ratio - INTEGER
671 sk->sk_pacing_rate is set by TCP stack using a ratio applied
672 to current rate. (current_rate = cwnd * mss / srtt)
673 If TCP is in slow start, tcp_pacing_ss_ratio is applied
674 to let TCP probe for bigger speeds, assuming cwnd can be
675 doubled every other RTT.
678 tcp_pacing_ca_ratio - INTEGER
679 sk->sk_pacing_rate is set by TCP stack using a ratio applied
680 to current rate. (current_rate = cwnd * mss / srtt)
681 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
682 is applied to conservatively probe for bigger throughput.
685 tcp_tso_win_divisor - INTEGER
686 This allows control over what percentage of the congestion window
687 can be consumed by a single TSO frame.
688 The setting of this parameter is a choice between burstiness and
689 building larger TSO frames.
692 tcp_tw_reuse - INTEGER
693 Enable reuse of TIME-WAIT sockets for new connections when it is
694 safe from protocol viewpoint.
697 2 - enable for loopback traffic only
698 It should not be changed without advice/request of technical
702 tcp_window_scaling - BOOLEAN
703 Enable window scaling as defined in RFC1323.
705 tcp_wmem - vector of 3 INTEGERs: min, default, max
706 min: Amount of memory reserved for send buffers for TCP sockets.
707 Each TCP socket has rights to use it due to fact of its birth.
710 default: initial size of send buffer used by TCP sockets. This
711 value overrides net.core.wmem_default used by other protocols.
712 It is usually lower than net.core.wmem_default.
715 max: Maximal amount of memory allowed for automatically tuned
716 send buffers for TCP sockets. This value does not override
717 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
718 automatic tuning of that socket's send buffer size, in which case
719 this value is ignored.
720 Default: between 64K and 4MB, depending on RAM size.
722 tcp_notsent_lowat - UNSIGNED INTEGER
723 A TCP socket can control the amount of unsent bytes in its write queue,
724 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
725 reports POLLOUT events if the amount of unsent bytes is below a per
726 socket value, and if the write queue is not full. sendmsg() will
727 also not add new buffers if the limit is hit.
729 This global variable controls the amount of unsent data for
730 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
731 to the global variable has immediate effect.
733 Default: UINT_MAX (0xFFFFFFFF)
735 tcp_workaround_signed_windows - BOOLEAN
736 If set, assume no receipt of a window scaling option means the
737 remote TCP is broken and treats the window as a signed quantity.
738 If unset, assume the remote TCP is not broken even if we do
739 not receive a window scaling option from them.
742 tcp_thin_linear_timeouts - BOOLEAN
743 Enable dynamic triggering of linear timeouts for thin streams.
744 If set, a check is performed upon retransmission by timeout to
745 determine if the stream is thin (less than 4 packets in flight).
746 As long as the stream is found to be thin, up to 6 linear
747 timeouts may be performed before exponential backoff mode is
748 initiated. This improves retransmission latency for
749 non-aggressive thin streams, often found to be time-dependent.
750 For more information on thin streams, see
751 Documentation/networking/tcp-thin.txt
754 tcp_limit_output_bytes - INTEGER
755 Controls TCP Small Queue limit per tcp socket.
756 TCP bulk sender tends to increase packets in flight until it
757 gets losses notifications. With SNDBUF autotuning, this can
758 result in a large amount of packets queued on the local machine
759 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
760 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
761 limits the number of bytes on qdisc or device to reduce artificial
762 RTT/cwnd and reduce bufferbloat.
763 Default: 1048576 (16 * 65536)
765 tcp_challenge_ack_limit - INTEGER
766 Limits number of Challenge ACK sent per second, as recommended
767 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
772 udp_l3mdev_accept - BOOLEAN
773 Enabling this option allows a "global" bound socket to work
774 across L3 master domains (e.g., VRFs) with packets capable of
775 being received regardless of the L3 domain in which they
776 originated. Only valid when the kernel was compiled with
777 CONFIG_NET_L3_MASTER_DEV.
778 Default: 0 (disabled)
780 udp_mem - vector of 3 INTEGERs: min, pressure, max
781 Number of pages allowed for queueing by all UDP sockets.
783 min: Below this number of pages UDP is not bothered about its
784 memory appetite. When amount of memory allocated by UDP exceeds
785 this number, UDP starts to moderate memory usage.
787 pressure: This value was introduced to follow format of tcp_mem.
789 max: Number of pages allowed for queueing by all UDP sockets.
791 Default is calculated at boot time from amount of available memory.
793 udp_rmem_min - INTEGER
794 Minimal size of receive buffer used by UDP sockets in moderation.
795 Each UDP socket is able to use the size for receiving data, even if
796 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
799 udp_wmem_min - INTEGER
800 Minimal size of send buffer used by UDP sockets in moderation.
801 Each UDP socket is able to use the size for sending data, even if
802 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
807 raw_l3mdev_accept - BOOLEAN
808 Enabling this option allows a "global" bound socket to work
809 across L3 master domains (e.g., VRFs) with packets capable of
810 being received regardless of the L3 domain in which they
811 originated. Only valid when the kernel was compiled with
812 CONFIG_NET_L3_MASTER_DEV.
817 cipso_cache_enable - BOOLEAN
818 If set, enable additions to and lookups from the CIPSO label mapping
819 cache. If unset, additions are ignored and lookups always result in a
820 miss. However, regardless of the setting the cache is still
821 invalidated when required when means you can safely toggle this on and
822 off and the cache will always be "safe".
825 cipso_cache_bucket_size - INTEGER
826 The CIPSO label cache consists of a fixed size hash table with each
827 hash bucket containing a number of cache entries. This variable limits
828 the number of entries in each hash bucket; the larger the value the
829 more CIPSO label mappings that can be cached. When the number of
830 entries in a given hash bucket reaches this limit adding new entries
831 causes the oldest entry in the bucket to be removed to make room.
834 cipso_rbm_optfmt - BOOLEAN
835 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
836 the CIPSO draft specification (see Documentation/netlabel for details).
837 This means that when set the CIPSO tag will be padded with empty
838 categories in order to make the packet data 32-bit aligned.
841 cipso_rbm_structvalid - BOOLEAN
842 If set, do a very strict check of the CIPSO option when
843 ip_options_compile() is called. If unset, relax the checks done during
844 ip_options_compile(). Either way is "safe" as errors are caught else
845 where in the CIPSO processing code but setting this to 0 (False) should
846 result in less work (i.e. it should be faster) but could cause problems
847 with other implementations that require strict checking.
852 ip_local_port_range - 2 INTEGERS
853 Defines the local port range that is used by TCP and UDP to
854 choose the local port. The first number is the first, the
855 second the last local port number.
856 If possible, it is better these numbers have different parity.
857 (one even and one odd values)
858 The default values are 32768 and 60999 respectively.
860 ip_local_reserved_ports - list of comma separated ranges
861 Specify the ports which are reserved for known third-party
862 applications. These ports will not be used by automatic port
863 assignments (e.g. when calling connect() or bind() with port
864 number 0). Explicit port allocation behavior is unchanged.
866 The format used for both input and output is a comma separated
867 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
868 10). Writing to the file will clear all previously reserved
869 ports and update the current list with the one given in the
872 Note that ip_local_port_range and ip_local_reserved_ports
873 settings are independent and both are considered by the kernel
874 when determining which ports are available for automatic port
877 You can reserve ports which are not in the current
878 ip_local_port_range, e.g.:
880 $ cat /proc/sys/net/ipv4/ip_local_port_range
882 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
885 although this is redundant. However such a setting is useful
886 if later the port range is changed to a value that will
887 include the reserved ports.
891 ip_unprivileged_port_start - INTEGER
892 This is a per-namespace sysctl. It defines the first
893 unprivileged port in the network namespace. Privileged ports
894 require root or CAP_NET_BIND_SERVICE in order to bind to them.
895 To disable all privileged ports, set this to 0. It may not
896 overlap with the ip_local_reserved_ports range.
900 ip_nonlocal_bind - BOOLEAN
901 If set, allows processes to bind() to non-local IP addresses,
902 which can be quite useful - but may break some applications.
906 If set non-zero, enables support for dynamic addresses.
907 If set to a non-zero value larger than 1, a kernel log
908 message will be printed when dynamic address rewriting
912 ip_early_demux - BOOLEAN
913 Optimize input packet processing down to one demux for
914 certain kinds of local sockets. Currently we only do this
915 for established TCP and connected UDP sockets.
917 It may add an additional cost for pure routing workloads that
918 reduces overall throughput, in such case you should disable it.
921 tcp_early_demux - BOOLEAN
922 Enable early demux for established TCP sockets.
925 udp_early_demux - BOOLEAN
926 Enable early demux for connected UDP sockets. Disable this if
927 your system could experience more unconnected load.
930 icmp_echo_ignore_all - BOOLEAN
931 If set non-zero, then the kernel will ignore all ICMP ECHO
935 icmp_echo_ignore_broadcasts - BOOLEAN
936 If set non-zero, then the kernel will ignore all ICMP ECHO and
937 TIMESTAMP requests sent to it via broadcast/multicast.
940 icmp_ratelimit - INTEGER
941 Limit the maximal rates for sending ICMP packets whose type matches
942 icmp_ratemask (see below) to specific targets.
943 0 to disable any limiting,
944 otherwise the minimal space between responses in milliseconds.
945 Note that another sysctl, icmp_msgs_per_sec limits the number
946 of ICMP packets sent on all targets.
949 icmp_msgs_per_sec - INTEGER
950 Limit maximal number of ICMP packets sent per second from this host.
951 Only messages whose type matches icmp_ratemask (see below) are
952 controlled by this limit.
955 icmp_msgs_burst - INTEGER
956 icmp_msgs_per_sec controls number of ICMP packets sent per second,
957 while icmp_msgs_burst controls the burst size of these packets.
960 icmp_ratemask - INTEGER
961 Mask made of ICMP types for which rates are being limited.
962 Significant bits: IHGFEDCBA9876543210
963 Default mask: 0000001100000011000 (6168)
965 Bit definitions (see include/linux/icmp.h):
967 3 Destination Unreachable *
972 C Parameter Problem *
977 H Address Mask Request
980 * These are rate limited by default (see default mask above)
982 icmp_ignore_bogus_error_responses - BOOLEAN
983 Some routers violate RFC1122 by sending bogus responses to broadcast
984 frames. Such violations are normally logged via a kernel warning.
985 If this is set to TRUE, the kernel will not give such warnings, which
986 will avoid log file clutter.
989 icmp_errors_use_inbound_ifaddr - BOOLEAN
991 If zero, icmp error messages are sent with the primary address of
992 the exiting interface.
994 If non-zero, the message will be sent with the primary address of
995 the interface that received the packet that caused the icmp error.
996 This is the behaviour network many administrators will expect from
997 a router. And it can make debugging complicated network layouts
1000 Note that if no primary address exists for the interface selected,
1001 then the primary address of the first non-loopback interface that
1002 has one will be used regardless of this setting.
1006 igmp_max_memberships - INTEGER
1007 Change the maximum number of multicast groups we can subscribe to.
1010 Theoretical maximum value is bounded by having to send a membership
1011 report in a single datagram (i.e. the report can't span multiple
1012 datagrams, or risk confusing the switch and leaving groups you don't
1015 The number of supported groups 'M' is bounded by the number of group
1016 report entries you can fit into a single datagram of 65535 bytes.
1018 M = 65536-sizeof (ip header)/(sizeof(Group record))
1020 Group records are variable length, with a minimum of 12 bytes.
1021 So net.ipv4.igmp_max_memberships should not be set higher than:
1023 (65536-24) / 12 = 5459
1025 The value 5459 assumes no IP header options, so in practice
1026 this number may be lower.
1028 igmp_max_msf - INTEGER
1029 Maximum number of addresses allowed in the source filter list for a
1034 Controls the IGMP query robustness variable (see RFC2236 8.1).
1035 Default: 2 (as specified by RFC2236 8.1)
1036 Minimum: 1 (as specified by RFC6636 4.5)
1038 force_igmp_version - INTEGER
1039 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1040 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1041 Present timer expires.
1042 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1043 receive IGMPv2/v3 query.
1044 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1045 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1046 3 - Enforce to use IGMP version 3. The same react with default 0.
1048 Note: this is not the same with force_mld_version because IGMPv3 RFC3376
1049 Security Considerations does not have clear description that we could
1050 ignore other version messages completely as MLDv2 RFC3810. So make
1051 this value as default 0 is recommended.
1053 conf/interface/* changes special settings per interface (where
1054 "interface" is the name of your network interface)
1056 conf/all/* is special, changes the settings for all interfaces
1058 log_martians - BOOLEAN
1059 Log packets with impossible addresses to kernel log.
1060 log_martians for the interface will be enabled if at least one of
1061 conf/{all,interface}/log_martians is set to TRUE,
1062 it will be disabled otherwise
1064 accept_redirects - BOOLEAN
1065 Accept ICMP redirect messages.
1066 accept_redirects for the interface will be enabled if:
1067 - both conf/{all,interface}/accept_redirects are TRUE in the case
1068 forwarding for the interface is enabled
1070 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1071 case forwarding for the interface is disabled
1072 accept_redirects for the interface will be disabled otherwise
1076 forwarding - BOOLEAN
1077 Enable IP forwarding on this interface. This controls whether packets
1078 received _on_ this interface can be forwarded.
1080 mc_forwarding - BOOLEAN
1081 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1082 and a multicast routing daemon is required.
1083 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1084 routing for the interface
1087 Integer value used to differentiate the devices by the medium they
1088 are attached to. Two devices can have different id values when
1089 the broadcast packets are received only on one of them.
1090 The default value 0 means that the device is the only interface
1091 to its medium, value of -1 means that medium is not known.
1093 Currently, it is used to change the proxy_arp behavior:
1094 the proxy_arp feature is enabled for packets forwarded between
1095 two devices attached to different media.
1099 proxy_arp for the interface will be enabled if at least one of
1100 conf/{all,interface}/proxy_arp is set to TRUE,
1101 it will be disabled otherwise
1103 proxy_arp_pvlan - BOOLEAN
1104 Private VLAN proxy arp.
1105 Basically allow proxy arp replies back to the same interface
1106 (from which the ARP request/solicitation was received).
1108 This is done to support (ethernet) switch features, like RFC
1109 3069, where the individual ports are NOT allowed to
1110 communicate with each other, but they are allowed to talk to
1111 the upstream router. As described in RFC 3069, it is possible
1112 to allow these hosts to communicate through the upstream
1113 router by proxy_arp'ing. Don't need to be used together with
1116 This technology is known by different names:
1117 In RFC 3069 it is called VLAN Aggregation.
1118 Cisco and Allied Telesyn call it Private VLAN.
1119 Hewlett-Packard call it Source-Port filtering or port-isolation.
1120 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1122 shared_media - BOOLEAN
1123 Send(router) or accept(host) RFC1620 shared media redirects.
1124 Overrides secure_redirects.
1125 shared_media for the interface will be enabled if at least one of
1126 conf/{all,interface}/shared_media is set to TRUE,
1127 it will be disabled otherwise
1130 secure_redirects - BOOLEAN
1131 Accept ICMP redirect messages only to gateways listed in the
1132 interface's current gateway list. Even if disabled, RFC1122 redirect
1134 Overridden by shared_media.
1135 secure_redirects for the interface will be enabled if at least one of
1136 conf/{all,interface}/secure_redirects is set to TRUE,
1137 it will be disabled otherwise
1140 send_redirects - BOOLEAN
1141 Send redirects, if router.
1142 send_redirects for the interface will be enabled if at least one of
1143 conf/{all,interface}/send_redirects is set to TRUE,
1144 it will be disabled otherwise
1147 bootp_relay - BOOLEAN
1148 Accept packets with source address 0.b.c.d destined
1149 not to this host as local ones. It is supposed, that
1150 BOOTP relay daemon will catch and forward such packets.
1151 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1154 Not Implemented Yet.
1156 accept_source_route - BOOLEAN
1157 Accept packets with SRR option.
1158 conf/all/accept_source_route must also be set to TRUE to accept packets
1159 with SRR option on the interface
1160 default TRUE (router)
1163 accept_local - BOOLEAN
1164 Accept packets with local source addresses. In combination with
1165 suitable routing, this can be used to direct packets between two
1166 local interfaces over the wire and have them accepted properly.
1169 route_localnet - BOOLEAN
1170 Do not consider loopback addresses as martian source or destination
1171 while routing. This enables the use of 127/8 for local routing purposes.
1175 0 - No source validation.
1176 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1177 Each incoming packet is tested against the FIB and if the interface
1178 is not the best reverse path the packet check will fail.
1179 By default failed packets are discarded.
1180 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1181 Each incoming packet's source address is also tested against the FIB
1182 and if the source address is not reachable via any interface
1183 the packet check will fail.
1185 Current recommended practice in RFC3704 is to enable strict mode
1186 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1187 or other complicated routing, then loose mode is recommended.
1189 The max value from conf/{all,interface}/rp_filter is used
1190 when doing source validation on the {interface}.
1192 Default value is 0. Note that some distributions enable it
1195 arp_filter - BOOLEAN
1196 1 - Allows you to have multiple network interfaces on the same
1197 subnet, and have the ARPs for each interface be answered
1198 based on whether or not the kernel would route a packet from
1199 the ARP'd IP out that interface (therefore you must use source
1200 based routing for this to work). In other words it allows control
1201 of which cards (usually 1) will respond to an arp request.
1203 0 - (default) The kernel can respond to arp requests with addresses
1204 from other interfaces. This may seem wrong but it usually makes
1205 sense, because it increases the chance of successful communication.
1206 IP addresses are owned by the complete host on Linux, not by
1207 particular interfaces. Only for more complex setups like load-
1208 balancing, does this behaviour cause problems.
1210 arp_filter for the interface will be enabled if at least one of
1211 conf/{all,interface}/arp_filter is set to TRUE,
1212 it will be disabled otherwise
1214 arp_announce - INTEGER
1215 Define different restriction levels for announcing the local
1216 source IP address from IP packets in ARP requests sent on
1218 0 - (default) Use any local address, configured on any interface
1219 1 - Try to avoid local addresses that are not in the target's
1220 subnet for this interface. This mode is useful when target
1221 hosts reachable via this interface require the source IP
1222 address in ARP requests to be part of their logical network
1223 configured on the receiving interface. When we generate the
1224 request we will check all our subnets that include the
1225 target IP and will preserve the source address if it is from
1226 such subnet. If there is no such subnet we select source
1227 address according to the rules for level 2.
1228 2 - Always use the best local address for this target.
1229 In this mode we ignore the source address in the IP packet
1230 and try to select local address that we prefer for talks with
1231 the target host. Such local address is selected by looking
1232 for primary IP addresses on all our subnets on the outgoing
1233 interface that include the target IP address. If no suitable
1234 local address is found we select the first local address
1235 we have on the outgoing interface or on all other interfaces,
1236 with the hope we will receive reply for our request and
1237 even sometimes no matter the source IP address we announce.
1239 The max value from conf/{all,interface}/arp_announce is used.
1241 Increasing the restriction level gives more chance for
1242 receiving answer from the resolved target while decreasing
1243 the level announces more valid sender's information.
1245 arp_ignore - INTEGER
1246 Define different modes for sending replies in response to
1247 received ARP requests that resolve local target IP addresses:
1248 0 - (default): reply for any local target IP address, configured
1250 1 - reply only if the target IP address is local address
1251 configured on the incoming interface
1252 2 - reply only if the target IP address is local address
1253 configured on the incoming interface and both with the
1254 sender's IP address are part from same subnet on this interface
1255 3 - do not reply for local addresses configured with scope host,
1256 only resolutions for global and link addresses are replied
1258 8 - do not reply for all local addresses
1260 The max value from conf/{all,interface}/arp_ignore is used
1261 when ARP request is received on the {interface}
1263 arp_notify - BOOLEAN
1264 Define mode for notification of address and device changes.
1265 0 - (default): do nothing
1266 1 - Generate gratuitous arp requests when device is brought up
1267 or hardware address changes.
1269 arp_accept - BOOLEAN
1270 Define behavior for gratuitous ARP frames who's IP is not
1271 already present in the ARP table:
1272 0 - don't create new entries in the ARP table
1273 1 - create new entries in the ARP table
1275 Both replies and requests type gratuitous arp will trigger the
1276 ARP table to be updated, if this setting is on.
1278 If the ARP table already contains the IP address of the
1279 gratuitous arp frame, the arp table will be updated regardless
1280 if this setting is on or off.
1282 mcast_solicit - INTEGER
1283 The maximum number of multicast probes in INCOMPLETE state,
1284 when the associated hardware address is unknown. Defaults
1287 ucast_solicit - INTEGER
1288 The maximum number of unicast probes in PROBE state, when
1289 the hardware address is being reconfirmed. Defaults to 3.
1291 app_solicit - INTEGER
1292 The maximum number of probes to send to the user space ARP daemon
1293 via netlink before dropping back to multicast probes (see
1294 mcast_resolicit). Defaults to 0.
1296 mcast_resolicit - INTEGER
1297 The maximum number of multicast probes after unicast and
1298 app probes in PROBE state. Defaults to 0.
1300 disable_policy - BOOLEAN
1301 Disable IPSEC policy (SPD) for this interface
1303 disable_xfrm - BOOLEAN
1304 Disable IPSEC encryption on this interface, whatever the policy
1306 igmpv2_unsolicited_report_interval - INTEGER
1307 The interval in milliseconds in which the next unsolicited
1308 IGMPv1 or IGMPv2 report retransmit will take place.
1309 Default: 10000 (10 seconds)
1311 igmpv3_unsolicited_report_interval - INTEGER
1312 The interval in milliseconds in which the next unsolicited
1313 IGMPv3 report retransmit will take place.
1314 Default: 1000 (1 seconds)
1316 promote_secondaries - BOOLEAN
1317 When a primary IP address is removed from this interface
1318 promote a corresponding secondary IP address instead of
1319 removing all the corresponding secondary IP addresses.
1321 drop_unicast_in_l2_multicast - BOOLEAN
1322 Drop any unicast IP packets that are received in link-layer
1323 multicast (or broadcast) frames.
1324 This behavior (for multicast) is actually a SHOULD in RFC
1325 1122, but is disabled by default for compatibility reasons.
1328 drop_gratuitous_arp - BOOLEAN
1329 Drop all gratuitous ARP frames, for example if there's a known
1330 good ARP proxy on the network and such frames need not be used
1331 (or in the case of 802.11, must not be used to prevent attacks.)
1336 Allows you to write a number, which can be used as required.
1339 xfrm4_gc_thresh - INTEGER
1340 (Obsolete since linux-4.14)
1341 The threshold at which we will start garbage collecting for IPv4
1342 destination cache entries. At twice this value the system will
1343 refuse new allocations.
1345 igmp_link_local_mcast_reports - BOOLEAN
1346 Enable IGMP reports for link local multicast groups in the
1351 kuznet@ms2.inr.ac.ru
1357 delon.nicolas@wanadoo.fr
1362 /proc/sys/net/ipv6/* Variables:
1364 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1365 apply to IPv6 [XXX?].
1367 bindv6only - BOOLEAN
1368 Default value for IPV6_V6ONLY socket option,
1369 which restricts use of the IPv6 socket to IPv6 communication
1371 TRUE: disable IPv4-mapped address feature
1372 FALSE: enable IPv4-mapped address feature
1374 Default: FALSE (as specified in RFC3493)
1376 flowlabel_consistency - BOOLEAN
1377 Protect the consistency (and unicity) of flow label.
1378 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1384 auto_flowlabels - INTEGER
1385 Automatically generate flow labels based on a flow hash of the
1386 packet. This allows intermediate devices, such as routers, to
1387 identify packet flows for mechanisms like Equal Cost Multipath
1388 Routing (see RFC 6438).
1389 0: automatic flow labels are completely disabled
1390 1: automatic flow labels are enabled by default, they can be
1391 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1393 2: automatic flow labels are allowed, they may be enabled on a
1394 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1395 3: automatic flow labels are enabled and enforced, they cannot
1396 be disabled by the socket option
1399 flowlabel_state_ranges - BOOLEAN
1400 Split the flow label number space into two ranges. 0-0x7FFFF is
1401 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1402 is reserved for stateless flow labels as described in RFC6437.
1407 flowlabel_reflect - BOOLEAN
1408 Automatically reflect the flow label. Needed for Path MTU
1409 Discovery to work with Equal Cost Multipath Routing in anycast
1410 environments. See RFC 7690 and:
1411 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1416 fib_multipath_hash_policy - INTEGER
1417 Controls which hash policy to use for multipath routes.
1418 Default: 0 (Layer 3)
1420 0 - Layer 3 (source and destination addresses plus flow label)
1421 1 - Layer 4 (standard 5-tuple)
1423 anycast_src_echo_reply - BOOLEAN
1424 Controls the use of anycast addresses as source addresses for ICMPv6
1430 idgen_delay - INTEGER
1431 Controls the delay in seconds after which time to retry
1432 privacy stable address generation if a DAD conflict is
1434 Default: 1 (as specified in RFC7217)
1436 idgen_retries - INTEGER
1437 Controls the number of retries to generate a stable privacy
1438 address if a DAD conflict is detected.
1439 Default: 3 (as specified in RFC7217)
1442 Controls the MLD query robustness variable (see RFC3810 9.1).
1443 Default: 2 (as specified by RFC3810 9.1)
1444 Minimum: 1 (as specified by RFC6636 4.5)
1446 max_dst_opts_number - INTEGER
1447 Maximum number of non-padding TLVs allowed in a Destination
1448 options extension header. If this value is less than zero
1449 then unknown options are disallowed and the number of known
1450 TLVs allowed is the absolute value of this number.
1453 max_hbh_opts_number - INTEGER
1454 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1455 options extension header. If this value is less than zero
1456 then unknown options are disallowed and the number of known
1457 TLVs allowed is the absolute value of this number.
1460 max_dst_opts_length - INTEGER
1461 Maximum length allowed for a Destination options extension
1463 Default: INT_MAX (unlimited)
1465 max_hbh_length - INTEGER
1466 Maximum length allowed for a Hop-by-Hop options extension
1468 Default: INT_MAX (unlimited)
1470 skip_notify_on_dev_down - BOOLEAN
1471 Controls whether an RTM_DELROUTE message is generated for routes
1472 removed when a device is taken down or deleted. IPv4 does not
1473 generate this message; IPv6 does by default. Setting this sysctl
1474 to true skips the message, making IPv4 and IPv6 on par in relying
1475 on userspace caches to track link events and evict routes.
1476 Default: false (generate message)
1480 ip6frag_high_thresh - INTEGER
1481 Maximum memory used to reassemble IPv6 fragments. When
1482 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1483 the fragment handler will toss packets until ip6frag_low_thresh
1486 ip6frag_low_thresh - INTEGER
1487 See ip6frag_high_thresh
1489 ip6frag_time - INTEGER
1490 Time in seconds to keep an IPv6 fragment in memory.
1492 IPv6 Segment Routing:
1494 seg6_flowlabel - INTEGER
1495 Controls the behaviour of computing the flowlabel of outer
1496 IPv6 header in case of SR T.encaps
1498 -1 set flowlabel to zero.
1499 0 copy flowlabel from Inner packet in case of Inner IPv6
1500 (Set flowlabel to 0 in case IPv4/L2)
1501 1 Compute the flowlabel using seg6_make_flowlabel()
1506 Change the interface-specific default settings.
1510 Change all the interface-specific settings.
1512 [XXX: Other special features than forwarding?]
1514 conf/all/forwarding - BOOLEAN
1515 Enable global IPv6 forwarding between all interfaces.
1517 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1518 to control which interfaces may forward packets and which not.
1520 This also sets all interfaces' Host/Router setting
1521 'forwarding' to the specified value. See below for details.
1523 This referred to as global forwarding.
1528 fwmark_reflect - BOOLEAN
1529 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1530 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1531 If unset, these packets have a fwmark of zero. If set, they have the
1532 fwmark of the packet they are replying to.
1536 Change special settings per interface.
1538 The functional behaviour for certain settings is different
1539 depending on whether local forwarding is enabled or not.
1542 Accept Router Advertisements; autoconfigure using them.
1544 It also determines whether or not to transmit Router
1545 Solicitations. If and only if the functional setting is to
1546 accept Router Advertisements, Router Solicitations will be
1549 Possible values are:
1550 0 Do not accept Router Advertisements.
1551 1 Accept Router Advertisements if forwarding is disabled.
1552 2 Overrule forwarding behaviour. Accept Router Advertisements
1553 even if forwarding is enabled.
1555 Functional default: enabled if local forwarding is disabled.
1556 disabled if local forwarding is enabled.
1558 accept_ra_defrtr - BOOLEAN
1559 Learn default router in Router Advertisement.
1561 Functional default: enabled if accept_ra is enabled.
1562 disabled if accept_ra is disabled.
1564 accept_ra_from_local - BOOLEAN
1565 Accept RA with source-address that is found on local machine
1566 if the RA is otherwise proper and able to be accepted.
1567 Default is to NOT accept these as it may be an un-intended
1571 enabled if accept_ra_from_local is enabled
1572 on a specific interface.
1573 disabled if accept_ra_from_local is disabled
1574 on a specific interface.
1576 accept_ra_min_hop_limit - INTEGER
1577 Minimum hop limit Information in Router Advertisement.
1579 Hop limit Information in Router Advertisement less than this
1580 variable shall be ignored.
1584 accept_ra_pinfo - BOOLEAN
1585 Learn Prefix Information in Router Advertisement.
1587 Functional default: enabled if accept_ra is enabled.
1588 disabled if accept_ra is disabled.
1590 accept_ra_rt_info_min_plen - INTEGER
1591 Minimum prefix length of Route Information in RA.
1593 Route Information w/ prefix smaller than this variable shall
1596 Functional default: 0 if accept_ra_rtr_pref is enabled.
1597 -1 if accept_ra_rtr_pref is disabled.
1599 accept_ra_rt_info_max_plen - INTEGER
1600 Maximum prefix length of Route Information in RA.
1602 Route Information w/ prefix larger than this variable shall
1605 Functional default: 0 if accept_ra_rtr_pref is enabled.
1606 -1 if accept_ra_rtr_pref is disabled.
1608 accept_ra_rtr_pref - BOOLEAN
1609 Accept Router Preference in RA.
1611 Functional default: enabled if accept_ra is enabled.
1612 disabled if accept_ra is disabled.
1614 accept_ra_mtu - BOOLEAN
1615 Apply the MTU value specified in RA option 5 (RFC4861). If
1616 disabled, the MTU specified in the RA will be ignored.
1618 Functional default: enabled if accept_ra is enabled.
1619 disabled if accept_ra is disabled.
1621 accept_redirects - BOOLEAN
1624 Functional default: enabled if local forwarding is disabled.
1625 disabled if local forwarding is enabled.
1627 accept_source_route - INTEGER
1628 Accept source routing (routing extension header).
1630 >= 0: Accept only routing header type 2.
1631 < 0: Do not accept routing header.
1636 Autoconfigure addresses using Prefix Information in Router
1639 Functional default: enabled if accept_ra_pinfo is enabled.
1640 disabled if accept_ra_pinfo is disabled.
1642 dad_transmits - INTEGER
1643 The amount of Duplicate Address Detection probes to send.
1646 forwarding - INTEGER
1647 Configure interface-specific Host/Router behaviour.
1649 Note: It is recommended to have the same setting on all
1650 interfaces; mixed router/host scenarios are rather uncommon.
1652 Possible values are:
1653 0 Forwarding disabled
1654 1 Forwarding enabled
1658 By default, Host behaviour is assumed. This means:
1660 1. IsRouter flag is not set in Neighbour Advertisements.
1661 2. If accept_ra is TRUE (default), transmit Router
1663 3. If accept_ra is TRUE (default), accept Router
1664 Advertisements (and do autoconfiguration).
1665 4. If accept_redirects is TRUE (default), accept Redirects.
1669 If local forwarding is enabled, Router behaviour is assumed.
1670 This means exactly the reverse from the above:
1672 1. IsRouter flag is set in Neighbour Advertisements.
1673 2. Router Solicitations are not sent unless accept_ra is 2.
1674 3. Router Advertisements are ignored unless accept_ra is 2.
1675 4. Redirects are ignored.
1677 Default: 0 (disabled) if global forwarding is disabled (default),
1678 otherwise 1 (enabled).
1681 Default Hop Limit to set.
1685 Default Maximum Transfer Unit
1686 Default: 1280 (IPv6 required minimum)
1688 ip_nonlocal_bind - BOOLEAN
1689 If set, allows processes to bind() to non-local IPv6 addresses,
1690 which can be quite useful - but may break some applications.
1693 router_probe_interval - INTEGER
1694 Minimum interval (in seconds) between Router Probing described
1699 router_solicitation_delay - INTEGER
1700 Number of seconds to wait after interface is brought up
1701 before sending Router Solicitations.
1704 router_solicitation_interval - INTEGER
1705 Number of seconds to wait between Router Solicitations.
1708 router_solicitations - INTEGER
1709 Number of Router Solicitations to send until assuming no
1710 routers are present.
1713 use_oif_addrs_only - BOOLEAN
1714 When enabled, the candidate source addresses for destinations
1715 routed via this interface are restricted to the set of addresses
1716 configured on this interface (vis. RFC 6724, section 4).
1720 use_tempaddr - INTEGER
1721 Preference for Privacy Extensions (RFC3041).
1722 <= 0 : disable Privacy Extensions
1723 == 1 : enable Privacy Extensions, but prefer public
1724 addresses over temporary addresses.
1725 > 1 : enable Privacy Extensions and prefer temporary
1726 addresses over public addresses.
1727 Default: 0 (for most devices)
1728 -1 (for point-to-point devices and loopback devices)
1730 temp_valid_lft - INTEGER
1731 valid lifetime (in seconds) for temporary addresses.
1732 Default: 604800 (7 days)
1734 temp_prefered_lft - INTEGER
1735 Preferred lifetime (in seconds) for temporary addresses.
1736 Default: 86400 (1 day)
1738 keep_addr_on_down - INTEGER
1739 Keep all IPv6 addresses on an interface down event. If set static
1740 global addresses with no expiration time are not flushed.
1745 Default: 0 (addresses are removed)
1747 max_desync_factor - INTEGER
1748 Maximum value for DESYNC_FACTOR, which is a random value
1749 that ensures that clients don't synchronize with each
1750 other and generate new addresses at exactly the same time.
1751 value is in seconds.
1754 regen_max_retry - INTEGER
1755 Number of attempts before give up attempting to generate
1756 valid temporary addresses.
1759 max_addresses - INTEGER
1760 Maximum number of autoconfigured addresses per interface. Setting
1761 to zero disables the limitation. It is not recommended to set this
1762 value too large (or to zero) because it would be an easy way to
1763 crash the kernel by allowing too many addresses to be created.
1766 disable_ipv6 - BOOLEAN
1767 Disable IPv6 operation. If accept_dad is set to 2, this value
1768 will be dynamically set to TRUE if DAD fails for the link-local
1770 Default: FALSE (enable IPv6 operation)
1772 When this value is changed from 1 to 0 (IPv6 is being enabled),
1773 it will dynamically create a link-local address on the given
1774 interface and start Duplicate Address Detection, if necessary.
1776 When this value is changed from 0 to 1 (IPv6 is being disabled),
1777 it will dynamically delete all addresses and routes on the given
1778 interface. From now on it will not possible to add addresses/routes
1779 to the selected interface.
1781 accept_dad - INTEGER
1782 Whether to accept DAD (Duplicate Address Detection).
1784 1: Enable DAD (default)
1785 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1786 link-local address has been found.
1788 DAD operation and mode on a given interface will be selected according
1789 to the maximum value of conf/{all,interface}/accept_dad.
1791 force_tllao - BOOLEAN
1792 Enable sending the target link-layer address option even when
1793 responding to a unicast neighbor solicitation.
1796 Quoting from RFC 2461, section 4.4, Target link-layer address:
1798 "The option MUST be included for multicast solicitations in order to
1799 avoid infinite Neighbor Solicitation "recursion" when the peer node
1800 does not have a cache entry to return a Neighbor Advertisements
1801 message. When responding to unicast solicitations, the option can be
1802 omitted since the sender of the solicitation has the correct link-
1803 layer address; otherwise it would not have be able to send the unicast
1804 solicitation in the first place. However, including the link-layer
1805 address in this case adds little overhead and eliminates a potential
1806 race condition where the sender deletes the cached link-layer address
1807 prior to receiving a response to a previous solicitation."
1809 ndisc_notify - BOOLEAN
1810 Define mode for notification of address and device changes.
1811 0 - (default): do nothing
1812 1 - Generate unsolicited neighbour advertisements when device is brought
1813 up or hardware address changes.
1815 ndisc_tclass - INTEGER
1816 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
1817 Discovery (Router Solicitation, Router Advertisement, Neighbor
1818 Solicitation, Neighbor Advertisement, Redirect) messages.
1819 These 8 bits can be interpreted as 6 high order bits holding the DSCP
1820 value and 2 low order bits representing ECN (which you probably want
1824 mldv1_unsolicited_report_interval - INTEGER
1825 The interval in milliseconds in which the next unsolicited
1826 MLDv1 report retransmit will take place.
1827 Default: 10000 (10 seconds)
1829 mldv2_unsolicited_report_interval - INTEGER
1830 The interval in milliseconds in which the next unsolicited
1831 MLDv2 report retransmit will take place.
1832 Default: 1000 (1 second)
1834 force_mld_version - INTEGER
1835 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1836 1 - Enforce to use MLD version 1
1837 2 - Enforce to use MLD version 2
1839 suppress_frag_ndisc - INTEGER
1840 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1841 with IPv6 Neighbor Discovery) behavior:
1842 1 - (default) discard fragmented neighbor discovery packets
1843 0 - allow fragmented neighbor discovery packets
1845 optimistic_dad - BOOLEAN
1846 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1847 0: disabled (default)
1850 Optimistic Duplicate Address Detection for the interface will be enabled
1851 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
1852 it will be disabled otherwise.
1854 use_optimistic - BOOLEAN
1855 If enabled, do not classify optimistic addresses as deprecated during
1856 source address selection. Preferred addresses will still be chosen
1857 before optimistic addresses, subject to other ranking in the source
1858 address selection algorithm.
1859 0: disabled (default)
1862 This will be enabled if at least one of
1863 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
1865 stable_secret - IPv6 address
1866 This IPv6 address will be used as a secret to generate IPv6
1867 addresses for link-local addresses and autoconfigured
1868 ones. All addresses generated after setting this secret will
1869 be stable privacy ones by default. This can be changed via the
1870 addrgenmode ip-link. conf/default/stable_secret is used as the
1871 secret for the namespace, the interface specific ones can
1872 overwrite that. Writes to conf/all/stable_secret are refused.
1874 It is recommended to generate this secret during installation
1875 of a system and keep it stable after that.
1877 By default the stable secret is unset.
1879 addr_gen_mode - INTEGER
1880 Defines how link-local and autoconf addresses are generated.
1882 0: generate address based on EUI64 (default)
1883 1: do no generate a link-local address, use EUI64 for addresses generated
1885 2: generate stable privacy addresses, using the secret from
1886 stable_secret (RFC7217)
1887 3: generate stable privacy addresses, using a random secret if unset
1889 drop_unicast_in_l2_multicast - BOOLEAN
1890 Drop any unicast IPv6 packets that are received in link-layer
1891 multicast (or broadcast) frames.
1893 By default this is turned off.
1895 drop_unsolicited_na - BOOLEAN
1896 Drop all unsolicited neighbor advertisements, for example if there's
1897 a known good NA proxy on the network and such frames need not be used
1898 (or in the case of 802.11, must not be used to prevent attacks.)
1900 By default this is turned off.
1902 enhanced_dad - BOOLEAN
1903 Include a nonce option in the IPv6 neighbor solicitation messages used for
1904 duplicate address detection per RFC7527. A received DAD NS will only signal
1905 a duplicate address if the nonce is different. This avoids any false
1906 detection of duplicates due to loopback of the NS messages that we send.
1907 The nonce option will be sent on an interface unless both of
1908 conf/{all,interface}/enhanced_dad are set to FALSE.
1913 Limit the maximal rates for sending ICMPv6 packets.
1914 0 to disable any limiting,
1915 otherwise the minimal space between responses in milliseconds.
1918 echo_ignore_all - BOOLEAN
1919 If set non-zero, then the kernel will ignore all ICMP ECHO
1920 requests sent to it over the IPv6 protocol.
1923 xfrm6_gc_thresh - INTEGER
1924 (Obsolete since linux-4.14)
1925 The threshold at which we will start garbage collecting for IPv6
1926 destination cache entries. At twice this value the system will
1927 refuse new allocations.
1931 Pekka Savola <pekkas@netcore.fi>
1932 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1935 /proc/sys/net/bridge/* Variables:
1937 bridge-nf-call-arptables - BOOLEAN
1938 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1942 bridge-nf-call-iptables - BOOLEAN
1943 1 : pass bridged IPv4 traffic to iptables' chains.
1947 bridge-nf-call-ip6tables - BOOLEAN
1948 1 : pass bridged IPv6 traffic to ip6tables' chains.
1952 bridge-nf-filter-vlan-tagged - BOOLEAN
1953 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1957 bridge-nf-filter-pppoe-tagged - BOOLEAN
1958 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1962 bridge-nf-pass-vlan-input-dev - BOOLEAN
1963 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1964 interface on the bridge and set the netfilter input device to the vlan.
1965 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1966 target work with vlan-on-top-of-bridge interfaces. When no matching
1967 vlan interface is found, or this switch is off, the input device is
1968 set to the bridge interface.
1969 0: disable bridge netfilter vlan interface lookup.
1972 proc/sys/net/sctp/* Variables:
1974 addip_enable - BOOLEAN
1975 Enable or disable extension of Dynamic Address Reconfiguration
1976 (ADD-IP) functionality specified in RFC5061. This extension provides
1977 the ability to dynamically add and remove new addresses for the SCTP
1980 1: Enable extension.
1982 0: Disable extension.
1987 Enable or disable pf (pf is short for potentially failed) state. A value
1988 of pf_retrans > path_max_retrans also disables pf state. That is, one of
1989 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
1990 Since pf_retrans and path_max_retrans can be changed by userspace
1991 application, sometimes user expects to disable pf state by the value of
1992 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
1993 or path_max_retrans is changed by the user application, this pf state is
1994 enabled. As such, it is necessary to add this to dynamically enable
1995 and disable pf state. See:
1996 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2005 addip_noauth_enable - BOOLEAN
2006 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2007 authentication to protect the operations of adding or removing new
2008 addresses. This requirement is mandated so that unauthorized hosts
2009 would not be able to hijack associations. However, older
2010 implementations may not have implemented this requirement while
2011 allowing the ADD-IP extension. For reasons of interoperability,
2012 we provide this variable to control the enforcement of the
2013 authentication requirement.
2015 1: Allow ADD-IP extension to be used without authentication. This
2016 should only be set in a closed environment for interoperability
2017 with older implementations.
2019 0: Enforce the authentication requirement
2023 auth_enable - BOOLEAN
2024 Enable or disable Authenticated Chunks extension. This extension
2025 provides the ability to send and receive authenticated chunks and is
2026 required for secure operation of Dynamic Address Reconfiguration
2029 1: Enable this extension.
2030 0: Disable this extension.
2034 prsctp_enable - BOOLEAN
2035 Enable or disable the Partial Reliability extension (RFC3758) which
2036 is used to notify peers that a given DATA should no longer be expected.
2044 The limit of the number of new packets that can be initially sent. It
2045 controls how bursty the generated traffic can be.
2049 association_max_retrans - INTEGER
2050 Set the maximum number for retransmissions that an association can
2051 attempt deciding that the remote end is unreachable. If this value
2052 is exceeded, the association is terminated.
2056 max_init_retransmits - INTEGER
2057 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2058 that an association will attempt before declaring the destination
2059 unreachable and terminating.
2063 path_max_retrans - INTEGER
2064 The maximum number of retransmissions that will be attempted on a given
2065 path. Once this threshold is exceeded, the path is considered
2066 unreachable, and new traffic will use a different path when the
2067 association is multihomed.
2071 pf_retrans - INTEGER
2072 The number of retransmissions that will be attempted on a given path
2073 before traffic is redirected to an alternate transport (should one
2074 exist). Note this is distinct from path_max_retrans, as a path that
2075 passes the pf_retrans threshold can still be used. Its only
2076 deprioritized when a transmission path is selected by the stack. This
2077 setting is primarily used to enable fast failover mechanisms without
2078 having to reduce path_max_retrans to a very low value. See:
2079 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2080 for details. Note also that a value of pf_retrans > path_max_retrans
2081 disables this feature. Since both pf_retrans and path_max_retrans can
2082 be changed by userspace application, a variable pf_enable is used to
2087 rto_initial - INTEGER
2088 The initial round trip timeout value in milliseconds that will be used
2089 in calculating round trip times. This is the initial time interval
2090 for retransmissions.
2095 The maximum value (in milliseconds) of the round trip timeout. This
2096 is the largest time interval that can elapse between retransmissions.
2101 The minimum value (in milliseconds) of the round trip timeout. This
2102 is the smallest time interval the can elapse between retransmissions.
2106 hb_interval - INTEGER
2107 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2108 are sent at the specified interval on idle paths to probe the state of
2109 a given path between 2 associations.
2113 sack_timeout - INTEGER
2114 The amount of time (in milliseconds) that the implementation will wait
2119 valid_cookie_life - INTEGER
2120 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2121 is used during association establishment.
2125 cookie_preserve_enable - BOOLEAN
2126 Enable or disable the ability to extend the lifetime of the SCTP cookie
2127 that is used during the establishment phase of SCTP association
2129 1: Enable cookie lifetime extension.
2134 cookie_hmac_alg - STRING
2135 Select the hmac algorithm used when generating the cookie value sent by
2136 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2141 Ability to assign md5 or sha1 as the selected alg is predicated on the
2142 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2143 CONFIG_CRYPTO_SHA1).
2145 Default: Dependent on configuration. MD5 if available, else SHA1 if
2146 available, else none.
2148 rcvbuf_policy - INTEGER
2149 Determines if the receive buffer is attributed to the socket or to
2150 association. SCTP supports the capability to create multiple
2151 associations on a single socket. When using this capability, it is
2152 possible that a single stalled association that's buffering a lot
2153 of data may block other associations from delivering their data by
2154 consuming all of the receive buffer space. To work around this,
2155 the rcvbuf_policy could be set to attribute the receiver buffer space
2156 to each association instead of the socket. This prevents the described
2159 1: rcvbuf space is per association
2160 0: rcvbuf space is per socket
2164 sndbuf_policy - INTEGER
2165 Similar to rcvbuf_policy above, this applies to send buffer space.
2167 1: Send buffer is tracked per association
2168 0: Send buffer is tracked per socket.
2172 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2173 Number of pages allowed for queueing by all SCTP sockets.
2175 min: Below this number of pages SCTP is not bothered about its
2176 memory appetite. When amount of memory allocated by SCTP exceeds
2177 this number, SCTP starts to moderate memory usage.
2179 pressure: This value was introduced to follow format of tcp_mem.
2181 max: Number of pages allowed for queueing by all SCTP sockets.
2183 Default is calculated at boot time from amount of available memory.
2185 sctp_rmem - vector of 3 INTEGERs: min, default, max
2186 Only the first value ("min") is used, "default" and "max" are
2189 min: Minimal size of receive buffer used by SCTP socket.
2190 It is guaranteed to each SCTP socket (but not association) even
2191 under moderate memory pressure.
2195 sctp_wmem - vector of 3 INTEGERs: min, default, max
2196 Currently this tunable has no effect.
2198 addr_scope_policy - INTEGER
2199 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2201 0 - Disable IPv4 address scoping
2202 1 - Enable IPv4 address scoping
2203 2 - Follow draft but allow IPv4 private addresses
2204 3 - Follow draft but allow IPv4 link local addresses
2209 /proc/sys/net/core/*
2210 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
2213 /proc/sys/net/unix/*
2214 max_dgram_qlen - INTEGER
2215 The maximum length of dgram socket receive queue