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 fib_sync_mem - UNSIGNED INTEGER
85 Amount of dirty memory from fib entries that can be backlogged before
86 synchronize_rcu is forced.
87 Default: 512kB Minimum: 64kB Maximum: 64MB
89 ip_forward_update_priority - INTEGER
90 Whether to update SKB priority from "TOS" field in IPv4 header after it
91 is forwarded. The new SKB priority is mapped from TOS field value
92 according to an rt_tos2priority table (see e.g. man tc-prio).
93 Default: 1 (Update priority.)
95 0 - Do not update priority.
98 route/max_size - INTEGER
99 Maximum number of routes allowed in the kernel. Increase
100 this when using large numbers of interfaces and/or routes.
101 From linux kernel 3.6 onwards, this is deprecated for ipv4
102 as route cache is no longer used.
104 neigh/default/gc_thresh1 - INTEGER
105 Minimum number of entries to keep. Garbage collector will not
106 purge entries if there are fewer than this number.
109 neigh/default/gc_thresh2 - INTEGER
110 Threshold when garbage collector becomes more aggressive about
111 purging entries. Entries older than 5 seconds will be cleared
112 when over this number.
115 neigh/default/gc_thresh3 - INTEGER
116 Maximum number of non-PERMANENT neighbor entries allowed. Increase
117 this when using large numbers of interfaces and when communicating
118 with large numbers of directly-connected peers.
121 neigh/default/unres_qlen_bytes - INTEGER
122 The maximum number of bytes which may be used by packets
123 queued for each unresolved address by other network layers.
125 Setting negative value is meaningless and will return error.
126 Default: SK_WMEM_MAX, (same as net.core.wmem_default).
127 Exact value depends on architecture and kernel options,
128 but should be enough to allow queuing 256 packets
131 neigh/default/unres_qlen - INTEGER
132 The maximum number of packets which may be queued for each
133 unresolved address by other network layers.
134 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
135 Prior to linux 3.3, the default value is 3 which may cause
136 unexpected packet loss. The current default value is calculated
137 according to default value of unres_qlen_bytes and true size of
141 mtu_expires - INTEGER
142 Time, in seconds, that cached PMTU information is kept.
144 min_adv_mss - INTEGER
145 The advertised MSS depends on the first hop route MTU, but will
146 never be lower than this setting.
150 ipfrag_high_thresh - LONG INTEGER
151 Maximum memory used to reassemble IP fragments.
153 ipfrag_low_thresh - LONG INTEGER
154 (Obsolete since linux-4.17)
155 Maximum memory used to reassemble IP fragments before the kernel
156 begins to remove incomplete fragment queues to free up resources.
157 The kernel still accepts new fragments for defragmentation.
159 ipfrag_time - INTEGER
160 Time in seconds to keep an IP fragment in memory.
162 ipfrag_max_dist - INTEGER
163 ipfrag_max_dist is a non-negative integer value which defines the
164 maximum "disorder" which is allowed among fragments which share a
165 common IP source address. Note that reordering of packets is
166 not unusual, but if a large number of fragments arrive from a source
167 IP address while a particular fragment queue remains incomplete, it
168 probably indicates that one or more fragments belonging to that queue
169 have been lost. When ipfrag_max_dist is positive, an additional check
170 is done on fragments before they are added to a reassembly queue - if
171 ipfrag_max_dist (or more) fragments have arrived from a particular IP
172 address between additions to any IP fragment queue using that source
173 address, it's presumed that one or more fragments in the queue are
174 lost. The existing fragment queue will be dropped, and a new one
175 started. An ipfrag_max_dist value of zero disables this check.
177 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
178 result in unnecessarily dropping fragment queues when normal
179 reordering of packets occurs, which could lead to poor application
180 performance. Using a very large value, e.g. 50000, increases the
181 likelihood of incorrectly reassembling IP fragments that originate
182 from different IP datagrams, which could result in data corruption.
187 inet_peer_threshold - INTEGER
188 The approximate size of the storage. Starting from this threshold
189 entries will be thrown aggressively. This threshold also determines
190 entries' time-to-live and time intervals between garbage collection
191 passes. More entries, less time-to-live, less GC interval.
193 inet_peer_minttl - INTEGER
194 Minimum time-to-live of entries. Should be enough to cover fragment
195 time-to-live on the reassembling side. This minimum time-to-live is
196 guaranteed if the pool size is less than inet_peer_threshold.
199 inet_peer_maxttl - INTEGER
200 Maximum time-to-live of entries. Unused entries will expire after
201 this period of time if there is no memory pressure on the pool (i.e.
202 when the number of entries in the pool is very small).
208 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
209 Defaults to 128. See also tcp_max_syn_backlog for additional tuning
212 tcp_abort_on_overflow - BOOLEAN
213 If listening service is too slow to accept new connections,
214 reset them. Default state is FALSE. It means that if overflow
215 occurred due to a burst, connection will recover. Enable this
216 option _only_ if you are really sure that listening daemon
217 cannot be tuned to accept connections faster. Enabling this
218 option can harm clients of your server.
220 tcp_adv_win_scale - INTEGER
221 Count buffering overhead as bytes/2^tcp_adv_win_scale
222 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
224 Possible values are [-31, 31], inclusive.
227 tcp_allowed_congestion_control - STRING
228 Show/set the congestion control choices available to non-privileged
229 processes. The list is a subset of those listed in
230 tcp_available_congestion_control.
231 Default is "reno" and the default setting (tcp_congestion_control).
233 tcp_app_win - INTEGER
234 Reserve max(window/2^tcp_app_win, mss) of window for application
235 buffer. Value 0 is special, it means that nothing is reserved.
238 tcp_autocorking - BOOLEAN
239 Enable TCP auto corking :
240 When applications do consecutive small write()/sendmsg() system calls,
241 we try to coalesce these small writes as much as possible, to lower
242 total amount of sent packets. This is done if at least one prior
243 packet for the flow is waiting in Qdisc queues or device transmit
244 queue. Applications can still use TCP_CORK for optimal behavior
245 when they know how/when to uncork their sockets.
248 tcp_available_congestion_control - STRING
249 Shows the available congestion control choices that are registered.
250 More congestion control algorithms may be available as modules,
253 tcp_base_mss - INTEGER
254 The initial value of search_low to be used by the packetization layer
255 Path MTU discovery (MTU probing). If MTU probing is enabled,
256 this is the initial MSS used by the connection.
258 tcp_congestion_control - STRING
259 Set the congestion control algorithm to be used for new
260 connections. The algorithm "reno" is always available, but
261 additional choices may be available based on kernel configuration.
262 Default is set as part of kernel configuration.
263 For passive connections, the listener congestion control choice
265 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
268 Allows TCP to send "duplicate" SACKs.
270 tcp_early_retrans - INTEGER
271 Tail loss probe (TLP) converts RTOs occurring due to tail
272 losses into fast recovery (draft-ietf-tcpm-rack). Note that
273 TLP requires RACK to function properly (see tcp_recovery below)
280 Control use of Explicit Congestion Notification (ECN) by TCP.
281 ECN is used only when both ends of the TCP connection indicate
282 support for it. This feature is useful in avoiding losses due
283 to congestion by allowing supporting routers to signal
284 congestion before having to drop packets.
286 0 Disable ECN. Neither initiate nor accept ECN.
287 1 Enable ECN when requested by incoming connections and
288 also request ECN on outgoing connection attempts.
289 2 Enable ECN when requested by incoming connections
290 but do not request ECN on outgoing connections.
293 tcp_ecn_fallback - BOOLEAN
294 If the kernel detects that ECN connection misbehaves, enable fall
295 back to non-ECN. Currently, this knob implements the fallback
296 from RFC3168, section 6.1.1.1., but we reserve that in future,
297 additional detection mechanisms could be implemented under this
298 knob. The value is not used, if tcp_ecn or per route (or congestion
299 control) ECN settings are disabled.
300 Default: 1 (fallback enabled)
303 This is a legacy option, it has no effect anymore.
305 tcp_fin_timeout - INTEGER
306 The length of time an orphaned (no longer referenced by any
307 application) connection will remain in the FIN_WAIT_2 state
308 before it is aborted at the local end. While a perfectly
309 valid "receive only" state for an un-orphaned connection, an
310 orphaned connection in FIN_WAIT_2 state could otherwise wait
311 forever for the remote to close its end of the connection.
316 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
317 F-RTO is an enhanced recovery algorithm for TCP retransmission
318 timeouts. It is particularly beneficial in networks where the
319 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
320 modification. It does not require any support from the peer.
322 By default it's enabled with a non-zero value. 0 disables F-RTO.
324 tcp_fwmark_accept - BOOLEAN
325 If set, incoming connections to listening sockets that do not have a
326 socket mark will set the mark of the accepting socket to the fwmark of
327 the incoming SYN packet. This will cause all packets on that connection
328 (starting from the first SYNACK) to be sent with that fwmark. The
329 listening socket's mark is unchanged. Listening sockets that already
330 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
335 tcp_invalid_ratelimit - INTEGER
336 Limit the maximal rate for sending duplicate acknowledgments
337 in response to incoming TCP packets that are for an existing
338 connection but that are invalid due to any of these reasons:
340 (a) out-of-window sequence number,
341 (b) out-of-window acknowledgment number, or
342 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
344 This can help mitigate simple "ack loop" DoS attacks, wherein
345 a buggy or malicious middlebox or man-in-the-middle can
346 rewrite TCP header fields in manner that causes each endpoint
347 to think that the other is sending invalid TCP segments, thus
348 causing each side to send an unterminating stream of duplicate
349 acknowledgments for invalid segments.
351 Using 0 disables rate-limiting of dupacks in response to
352 invalid segments; otherwise this value specifies the minimal
353 space between sending such dupacks, in milliseconds.
355 Default: 500 (milliseconds).
357 tcp_keepalive_time - INTEGER
358 How often TCP sends out keepalive messages when keepalive is enabled.
361 tcp_keepalive_probes - INTEGER
362 How many keepalive probes TCP sends out, until it decides that the
363 connection is broken. Default value: 9.
365 tcp_keepalive_intvl - INTEGER
366 How frequently the probes are send out. Multiplied by
367 tcp_keepalive_probes it is time to kill not responding connection,
368 after probes started. Default value: 75sec i.e. connection
369 will be aborted after ~11 minutes of retries.
371 tcp_l3mdev_accept - BOOLEAN
372 Enables child sockets to inherit the L3 master device index.
373 Enabling this option allows a "global" listen socket to work
374 across L3 master domains (e.g., VRFs) with connected sockets
375 derived from the listen socket to be bound to the L3 domain in
376 which the packets originated. Only valid when the kernel was
377 compiled with CONFIG_NET_L3_MASTER_DEV.
378 Default: 0 (disabled)
380 tcp_low_latency - BOOLEAN
381 This is a legacy option, it has no effect anymore.
383 tcp_max_orphans - INTEGER
384 Maximal number of TCP sockets not attached to any user file handle,
385 held by system. If this number is exceeded orphaned connections are
386 reset immediately and warning is printed. This limit exists
387 only to prevent simple DoS attacks, you _must_ not rely on this
388 or lower the limit artificially, but rather increase it
389 (probably, after increasing installed memory),
390 if network conditions require more than default value,
391 and tune network services to linger and kill such states
392 more aggressively. Let me to remind again: each orphan eats
393 up to ~64K of unswappable memory.
395 tcp_max_syn_backlog - INTEGER
396 Maximal number of remembered connection requests, which have not
397 received an acknowledgment from connecting client.
398 The minimal value is 128 for low memory machines, and it will
399 increase in proportion to the memory of machine.
400 If server suffers from overload, try increasing this number.
402 tcp_max_tw_buckets - INTEGER
403 Maximal number of timewait sockets held by system simultaneously.
404 If this number is exceeded time-wait socket is immediately destroyed
405 and warning is printed. This limit exists only to prevent
406 simple DoS attacks, you _must_ not lower the limit artificially,
407 but rather increase it (probably, after increasing installed memory),
408 if network conditions require more than default value.
410 tcp_mem - vector of 3 INTEGERs: min, pressure, max
411 min: below this number of pages TCP is not bothered about its
414 pressure: when amount of memory allocated by TCP exceeds this number
415 of pages, TCP moderates its memory consumption and enters memory
416 pressure mode, which is exited when memory consumption falls
419 max: number of pages allowed for queueing by all TCP sockets.
421 Defaults are calculated at boot time from amount of available
424 tcp_min_rtt_wlen - INTEGER
425 The window length of the windowed min filter to track the minimum RTT.
426 A shorter window lets a flow more quickly pick up new (higher)
427 minimum RTT when it is moved to a longer path (e.g., due to traffic
428 engineering). A longer window makes the filter more resistant to RTT
429 inflations such as transient congestion. The unit is seconds.
430 Possible values: 0 - 86400 (1 day)
433 tcp_moderate_rcvbuf - BOOLEAN
434 If set, TCP performs receive buffer auto-tuning, attempting to
435 automatically size the buffer (no greater than tcp_rmem[2]) to
436 match the size required by the path for full throughput. Enabled by
439 tcp_mtu_probing - INTEGER
440 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
443 1 - Disabled by default, enabled when an ICMP black hole detected
444 2 - Always enabled, use initial MSS of tcp_base_mss.
446 tcp_probe_interval - UNSIGNED INTEGER
447 Controls how often to start TCP Packetization-Layer Path MTU
448 Discovery reprobe. The default is reprobing every 10 minutes as
451 tcp_probe_threshold - INTEGER
452 Controls when TCP Packetization-Layer Path MTU Discovery probing
453 will stop in respect to the width of search range in bytes. Default
456 tcp_no_metrics_save - BOOLEAN
457 By default, TCP saves various connection metrics in the route cache
458 when the connection closes, so that connections established in the
459 near future can use these to set initial conditions. Usually, this
460 increases overall performance, but may sometimes cause performance
461 degradation. If set, TCP will not cache metrics on closing
464 tcp_orphan_retries - INTEGER
465 This value influences the timeout of a locally closed TCP connection,
466 when RTO retransmissions remain unacknowledged.
467 See tcp_retries2 for more details.
469 The default value is 8.
470 If your machine is a loaded WEB server,
471 you should think about lowering this value, such sockets
472 may consume significant resources. Cf. tcp_max_orphans.
474 tcp_recovery - INTEGER
475 This value is a bitmap to enable various experimental loss recovery
478 RACK: 0x1 enables the RACK loss detection for fast detection of lost
479 retransmissions and tail drops. It also subsumes and disables
480 RFC6675 recovery for SACK connections.
481 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
482 RACK: 0x4 disables RACK's DUPACK threshold heuristic
486 tcp_reordering - INTEGER
487 Initial reordering level of packets in a TCP stream.
488 TCP stack can then dynamically adjust flow reordering level
489 between this initial value and tcp_max_reordering
492 tcp_max_reordering - INTEGER
493 Maximal reordering level of packets in a TCP stream.
494 300 is a fairly conservative value, but you might increase it
495 if paths are using per packet load balancing (like bonding rr mode)
498 tcp_retrans_collapse - BOOLEAN
499 Bug-to-bug compatibility with some broken printers.
500 On retransmit try to send bigger packets to work around bugs in
503 tcp_retries1 - INTEGER
504 This value influences the time, after which TCP decides, that
505 something is wrong due to unacknowledged RTO retransmissions,
506 and reports this suspicion to the network layer.
507 See tcp_retries2 for more details.
509 RFC 1122 recommends at least 3 retransmissions, which is the
512 tcp_retries2 - INTEGER
513 This value influences the timeout of an alive TCP connection,
514 when RTO retransmissions remain unacknowledged.
515 Given a value of N, a hypothetical TCP connection following
516 exponential backoff with an initial RTO of TCP_RTO_MIN would
517 retransmit N times before killing the connection at the (N+1)th RTO.
519 The default value of 15 yields a hypothetical timeout of 924.6
520 seconds and is a lower bound for the effective timeout.
521 TCP will effectively time out at the first RTO which exceeds the
522 hypothetical timeout.
524 RFC 1122 recommends at least 100 seconds for the timeout,
525 which corresponds to a value of at least 8.
527 tcp_rfc1337 - BOOLEAN
528 If set, the TCP stack behaves conforming to RFC1337. If unset,
529 we are not conforming to RFC, but prevent TCP TIME_WAIT
533 tcp_rmem - vector of 3 INTEGERs: min, default, max
534 min: Minimal size of receive buffer used by TCP sockets.
535 It is guaranteed to each TCP socket, even under moderate memory
539 default: initial size of receive buffer used by TCP sockets.
540 This value overrides net.core.rmem_default used by other protocols.
541 Default: 87380 bytes. This value results in window of 65535 with
542 default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
543 less for default tcp_app_win. See below about these variables.
545 max: maximal size of receive buffer allowed for automatically
546 selected receiver buffers for TCP socket. This value does not override
547 net.core.rmem_max. Calling setsockopt() with SO_RCVBUF disables
548 automatic tuning of that socket's receive buffer size, in which
549 case this value is ignored.
550 Default: between 87380B and 6MB, depending on RAM size.
553 Enable select acknowledgments (SACKS).
555 tcp_comp_sack_delay_ns - LONG INTEGER
556 TCP tries to reduce number of SACK sent, using a timer
557 based on 5% of SRTT, capped by this sysctl, in nano seconds.
558 The default is 1ms, based on TSO autosizing period.
560 Default : 1,000,000 ns (1 ms)
562 tcp_comp_sack_nr - INTEGER
563 Max number of SACK that can be compressed.
564 Using 0 disables SACK compression.
568 tcp_slow_start_after_idle - BOOLEAN
569 If set, provide RFC2861 behavior and time out the congestion
570 window after an idle period. An idle period is defined at
571 the current RTO. If unset, the congestion window will not
572 be timed out after an idle period.
576 Use the Host requirements interpretation of the TCP urgent pointer field.
577 Most hosts use the older BSD interpretation, so if you turn this on
578 Linux might not communicate correctly with them.
581 tcp_synack_retries - INTEGER
582 Number of times SYNACKs for a passive TCP connection attempt will
583 be retransmitted. Should not be higher than 255. Default value
584 is 5, which corresponds to 31seconds till the last retransmission
585 with the current initial RTO of 1second. With this the final timeout
586 for a passive TCP connection will happen after 63seconds.
588 tcp_syncookies - BOOLEAN
589 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
590 Send out syncookies when the syn backlog queue of a socket
591 overflows. This is to prevent against the common 'SYN flood attack'
594 Note, that syncookies is fallback facility.
595 It MUST NOT be used to help highly loaded servers to stand
596 against legal connection rate. If you see SYN flood warnings
597 in your logs, but investigation shows that they occur
598 because of overload with legal connections, you should tune
599 another parameters until this warning disappear.
600 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
602 syncookies seriously violate TCP protocol, do not allow
603 to use TCP extensions, can result in serious degradation
604 of some services (f.e. SMTP relaying), visible not by you,
605 but your clients and relays, contacting you. While you see
606 SYN flood warnings in logs not being really flooded, your server
607 is seriously misconfigured.
609 If you want to test which effects syncookies have to your
610 network connections you can set this knob to 2 to enable
611 unconditionally generation of syncookies.
613 tcp_fastopen - INTEGER
614 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
617 The client support is enabled by flag 0x1 (on by default). The client
618 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
619 rather than connect() to send data in SYN.
621 The server support is enabled by flag 0x2 (off by default). Then
622 either enable for all listeners with another flag (0x400) or
623 enable individual listeners via TCP_FASTOPEN socket option with
624 the option value being the length of the syn-data backlog.
626 The values (bitmap) are
627 0x1: (client) enables sending data in the opening SYN on the client.
628 0x2: (server) enables the server support, i.e., allowing data in
629 a SYN packet to be accepted and passed to the
630 application before 3-way handshake finishes.
631 0x4: (client) send data in the opening SYN regardless of cookie
632 availability and without a cookie option.
633 0x200: (server) accept data-in-SYN w/o any cookie option present.
634 0x400: (server) enable all listeners to support Fast Open by
635 default without explicit TCP_FASTOPEN socket option.
639 Note that that additional client or server features are only
640 effective if the basic support (0x1 and 0x2) are enabled respectively.
642 tcp_fastopen_blackhole_timeout_sec - INTEGER
643 Initial time period in second to disable Fastopen on active TCP sockets
644 when a TFO firewall blackhole issue happens.
645 This time period will grow exponentially when more blackhole issues
646 get detected right after Fastopen is re-enabled and will reset to
647 initial value when the blackhole issue goes away.
648 0 to disable the blackhole detection.
649 By default, it is set to 1hr.
651 tcp_syn_retries - INTEGER
652 Number of times initial SYNs for an active TCP connection attempt
653 will be retransmitted. Should not be higher than 127. Default value
654 is 6, which corresponds to 63seconds till the last retransmission
655 with the current initial RTO of 1second. With this the final timeout
656 for an active TCP connection attempt will happen after 127seconds.
658 tcp_timestamps - INTEGER
659 Enable timestamps as defined in RFC1323.
661 1: Enable timestamps as defined in RFC1323 and use random offset for
662 each connection rather than only using the current time.
663 2: Like 1, but without random offsets.
666 tcp_min_tso_segs - INTEGER
667 Minimal number of segments per TSO frame.
668 Since linux-3.12, TCP does an automatic sizing of TSO frames,
669 depending on flow rate, instead of filling 64Kbytes packets.
670 For specific usages, it's possible to force TCP to build big
671 TSO frames. Note that TCP stack might split too big TSO packets
672 if available window is too small.
675 tcp_pacing_ss_ratio - INTEGER
676 sk->sk_pacing_rate is set by TCP stack using a ratio applied
677 to current rate. (current_rate = cwnd * mss / srtt)
678 If TCP is in slow start, tcp_pacing_ss_ratio is applied
679 to let TCP probe for bigger speeds, assuming cwnd can be
680 doubled every other RTT.
683 tcp_pacing_ca_ratio - INTEGER
684 sk->sk_pacing_rate is set by TCP stack using a ratio applied
685 to current rate. (current_rate = cwnd * mss / srtt)
686 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
687 is applied to conservatively probe for bigger throughput.
690 tcp_tso_win_divisor - INTEGER
691 This allows control over what percentage of the congestion window
692 can be consumed by a single TSO frame.
693 The setting of this parameter is a choice between burstiness and
694 building larger TSO frames.
697 tcp_tw_reuse - INTEGER
698 Enable reuse of TIME-WAIT sockets for new connections when it is
699 safe from protocol viewpoint.
702 2 - enable for loopback traffic only
703 It should not be changed without advice/request of technical
707 tcp_window_scaling - BOOLEAN
708 Enable window scaling as defined in RFC1323.
710 tcp_wmem - vector of 3 INTEGERs: min, default, max
711 min: Amount of memory reserved for send buffers for TCP sockets.
712 Each TCP socket has rights to use it due to fact of its birth.
715 default: initial size of send buffer used by TCP sockets. This
716 value overrides net.core.wmem_default used by other protocols.
717 It is usually lower than net.core.wmem_default.
720 max: Maximal amount of memory allowed for automatically tuned
721 send buffers for TCP sockets. This value does not override
722 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
723 automatic tuning of that socket's send buffer size, in which case
724 this value is ignored.
725 Default: between 64K and 4MB, depending on RAM size.
727 tcp_notsent_lowat - UNSIGNED INTEGER
728 A TCP socket can control the amount of unsent bytes in its write queue,
729 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
730 reports POLLOUT events if the amount of unsent bytes is below a per
731 socket value, and if the write queue is not full. sendmsg() will
732 also not add new buffers if the limit is hit.
734 This global variable controls the amount of unsent data for
735 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
736 to the global variable has immediate effect.
738 Default: UINT_MAX (0xFFFFFFFF)
740 tcp_workaround_signed_windows - BOOLEAN
741 If set, assume no receipt of a window scaling option means the
742 remote TCP is broken and treats the window as a signed quantity.
743 If unset, assume the remote TCP is not broken even if we do
744 not receive a window scaling option from them.
747 tcp_thin_linear_timeouts - BOOLEAN
748 Enable dynamic triggering of linear timeouts for thin streams.
749 If set, a check is performed upon retransmission by timeout to
750 determine if the stream is thin (less than 4 packets in flight).
751 As long as the stream is found to be thin, up to 6 linear
752 timeouts may be performed before exponential backoff mode is
753 initiated. This improves retransmission latency for
754 non-aggressive thin streams, often found to be time-dependent.
755 For more information on thin streams, see
756 Documentation/networking/tcp-thin.txt
759 tcp_limit_output_bytes - INTEGER
760 Controls TCP Small Queue limit per tcp socket.
761 TCP bulk sender tends to increase packets in flight until it
762 gets losses notifications. With SNDBUF autotuning, this can
763 result in a large amount of packets queued on the local machine
764 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
765 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
766 limits the number of bytes on qdisc or device to reduce artificial
767 RTT/cwnd and reduce bufferbloat.
768 Default: 1048576 (16 * 65536)
770 tcp_challenge_ack_limit - INTEGER
771 Limits number of Challenge ACK sent per second, as recommended
772 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
777 udp_l3mdev_accept - BOOLEAN
778 Enabling this option allows a "global" bound socket to work
779 across L3 master domains (e.g., VRFs) with packets capable of
780 being received regardless of the L3 domain in which they
781 originated. Only valid when the kernel was compiled with
782 CONFIG_NET_L3_MASTER_DEV.
783 Default: 0 (disabled)
785 udp_mem - vector of 3 INTEGERs: min, pressure, max
786 Number of pages allowed for queueing by all UDP sockets.
788 min: Below this number of pages UDP is not bothered about its
789 memory appetite. When amount of memory allocated by UDP exceeds
790 this number, UDP starts to moderate memory usage.
792 pressure: This value was introduced to follow format of tcp_mem.
794 max: Number of pages allowed for queueing by all UDP sockets.
796 Default is calculated at boot time from amount of available memory.
798 udp_rmem_min - INTEGER
799 Minimal size of receive buffer used by UDP sockets in moderation.
800 Each UDP socket is able to use the size for receiving data, even if
801 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
804 udp_wmem_min - INTEGER
805 Minimal size of send buffer used by UDP sockets in moderation.
806 Each UDP socket is able to use the size for sending data, even if
807 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
812 raw_l3mdev_accept - BOOLEAN
813 Enabling this option allows a "global" bound socket to work
814 across L3 master domains (e.g., VRFs) with packets capable of
815 being received regardless of the L3 domain in which they
816 originated. Only valid when the kernel was compiled with
817 CONFIG_NET_L3_MASTER_DEV.
822 cipso_cache_enable - BOOLEAN
823 If set, enable additions to and lookups from the CIPSO label mapping
824 cache. If unset, additions are ignored and lookups always result in a
825 miss. However, regardless of the setting the cache is still
826 invalidated when required when means you can safely toggle this on and
827 off and the cache will always be "safe".
830 cipso_cache_bucket_size - INTEGER
831 The CIPSO label cache consists of a fixed size hash table with each
832 hash bucket containing a number of cache entries. This variable limits
833 the number of entries in each hash bucket; the larger the value the
834 more CIPSO label mappings that can be cached. When the number of
835 entries in a given hash bucket reaches this limit adding new entries
836 causes the oldest entry in the bucket to be removed to make room.
839 cipso_rbm_optfmt - BOOLEAN
840 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
841 the CIPSO draft specification (see Documentation/netlabel for details).
842 This means that when set the CIPSO tag will be padded with empty
843 categories in order to make the packet data 32-bit aligned.
846 cipso_rbm_structvalid - BOOLEAN
847 If set, do a very strict check of the CIPSO option when
848 ip_options_compile() is called. If unset, relax the checks done during
849 ip_options_compile(). Either way is "safe" as errors are caught else
850 where in the CIPSO processing code but setting this to 0 (False) should
851 result in less work (i.e. it should be faster) but could cause problems
852 with other implementations that require strict checking.
857 ip_local_port_range - 2 INTEGERS
858 Defines the local port range that is used by TCP and UDP to
859 choose the local port. The first number is the first, the
860 second the last local port number.
861 If possible, it is better these numbers have different parity.
862 (one even and one odd values)
863 The default values are 32768 and 60999 respectively.
865 ip_local_reserved_ports - list of comma separated ranges
866 Specify the ports which are reserved for known third-party
867 applications. These ports will not be used by automatic port
868 assignments (e.g. when calling connect() or bind() with port
869 number 0). Explicit port allocation behavior is unchanged.
871 The format used for both input and output is a comma separated
872 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
873 10). Writing to the file will clear all previously reserved
874 ports and update the current list with the one given in the
877 Note that ip_local_port_range and ip_local_reserved_ports
878 settings are independent and both are considered by the kernel
879 when determining which ports are available for automatic port
882 You can reserve ports which are not in the current
883 ip_local_port_range, e.g.:
885 $ cat /proc/sys/net/ipv4/ip_local_port_range
887 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
890 although this is redundant. However such a setting is useful
891 if later the port range is changed to a value that will
892 include the reserved ports.
896 ip_unprivileged_port_start - INTEGER
897 This is a per-namespace sysctl. It defines the first
898 unprivileged port in the network namespace. Privileged ports
899 require root or CAP_NET_BIND_SERVICE in order to bind to them.
900 To disable all privileged ports, set this to 0. It may not
901 overlap with the ip_local_reserved_ports range.
905 ip_nonlocal_bind - BOOLEAN
906 If set, allows processes to bind() to non-local IP addresses,
907 which can be quite useful - but may break some applications.
911 If set non-zero, enables support for dynamic addresses.
912 If set to a non-zero value larger than 1, a kernel log
913 message will be printed when dynamic address rewriting
917 ip_early_demux - BOOLEAN
918 Optimize input packet processing down to one demux for
919 certain kinds of local sockets. Currently we only do this
920 for established TCP and connected UDP sockets.
922 It may add an additional cost for pure routing workloads that
923 reduces overall throughput, in such case you should disable it.
926 tcp_early_demux - BOOLEAN
927 Enable early demux for established TCP sockets.
930 udp_early_demux - BOOLEAN
931 Enable early demux for connected UDP sockets. Disable this if
932 your system could experience more unconnected load.
935 icmp_echo_ignore_all - BOOLEAN
936 If set non-zero, then the kernel will ignore all ICMP ECHO
940 icmp_echo_ignore_broadcasts - BOOLEAN
941 If set non-zero, then the kernel will ignore all ICMP ECHO and
942 TIMESTAMP requests sent to it via broadcast/multicast.
945 icmp_ratelimit - INTEGER
946 Limit the maximal rates for sending ICMP packets whose type matches
947 icmp_ratemask (see below) to specific targets.
948 0 to disable any limiting,
949 otherwise the minimal space between responses in milliseconds.
950 Note that another sysctl, icmp_msgs_per_sec limits the number
951 of ICMP packets sent on all targets.
954 icmp_msgs_per_sec - INTEGER
955 Limit maximal number of ICMP packets sent per second from this host.
956 Only messages whose type matches icmp_ratemask (see below) are
957 controlled by this limit.
960 icmp_msgs_burst - INTEGER
961 icmp_msgs_per_sec controls number of ICMP packets sent per second,
962 while icmp_msgs_burst controls the burst size of these packets.
965 icmp_ratemask - INTEGER
966 Mask made of ICMP types for which rates are being limited.
967 Significant bits: IHGFEDCBA9876543210
968 Default mask: 0000001100000011000 (6168)
970 Bit definitions (see include/linux/icmp.h):
972 3 Destination Unreachable *
977 C Parameter Problem *
982 H Address Mask Request
985 * These are rate limited by default (see default mask above)
987 icmp_ignore_bogus_error_responses - BOOLEAN
988 Some routers violate RFC1122 by sending bogus responses to broadcast
989 frames. Such violations are normally logged via a kernel warning.
990 If this is set to TRUE, the kernel will not give such warnings, which
991 will avoid log file clutter.
994 icmp_errors_use_inbound_ifaddr - BOOLEAN
996 If zero, icmp error messages are sent with the primary address of
997 the exiting interface.
999 If non-zero, the message will be sent with the primary address of
1000 the interface that received the packet that caused the icmp error.
1001 This is the behaviour network many administrators will expect from
1002 a router. And it can make debugging complicated network layouts
1005 Note that if no primary address exists for the interface selected,
1006 then the primary address of the first non-loopback interface that
1007 has one will be used regardless of this setting.
1011 igmp_max_memberships - INTEGER
1012 Change the maximum number of multicast groups we can subscribe to.
1015 Theoretical maximum value is bounded by having to send a membership
1016 report in a single datagram (i.e. the report can't span multiple
1017 datagrams, or risk confusing the switch and leaving groups you don't
1020 The number of supported groups 'M' is bounded by the number of group
1021 report entries you can fit into a single datagram of 65535 bytes.
1023 M = 65536-sizeof (ip header)/(sizeof(Group record))
1025 Group records are variable length, with a minimum of 12 bytes.
1026 So net.ipv4.igmp_max_memberships should not be set higher than:
1028 (65536-24) / 12 = 5459
1030 The value 5459 assumes no IP header options, so in practice
1031 this number may be lower.
1033 igmp_max_msf - INTEGER
1034 Maximum number of addresses allowed in the source filter list for a
1039 Controls the IGMP query robustness variable (see RFC2236 8.1).
1040 Default: 2 (as specified by RFC2236 8.1)
1041 Minimum: 1 (as specified by RFC6636 4.5)
1043 force_igmp_version - INTEGER
1044 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1045 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1046 Present timer expires.
1047 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1048 receive IGMPv2/v3 query.
1049 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1050 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1051 3 - Enforce to use IGMP version 3. The same react with default 0.
1053 Note: this is not the same with force_mld_version because IGMPv3 RFC3376
1054 Security Considerations does not have clear description that we could
1055 ignore other version messages completely as MLDv2 RFC3810. So make
1056 this value as default 0 is recommended.
1058 conf/interface/* changes special settings per interface (where
1059 "interface" is the name of your network interface)
1061 conf/all/* is special, changes the settings for all interfaces
1063 log_martians - BOOLEAN
1064 Log packets with impossible addresses to kernel log.
1065 log_martians for the interface will be enabled if at least one of
1066 conf/{all,interface}/log_martians is set to TRUE,
1067 it will be disabled otherwise
1069 accept_redirects - BOOLEAN
1070 Accept ICMP redirect messages.
1071 accept_redirects for the interface will be enabled if:
1072 - both conf/{all,interface}/accept_redirects are TRUE in the case
1073 forwarding for the interface is enabled
1075 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1076 case forwarding for the interface is disabled
1077 accept_redirects for the interface will be disabled otherwise
1081 forwarding - BOOLEAN
1082 Enable IP forwarding on this interface. This controls whether packets
1083 received _on_ this interface can be forwarded.
1085 mc_forwarding - BOOLEAN
1086 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1087 and a multicast routing daemon is required.
1088 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1089 routing for the interface
1092 Integer value used to differentiate the devices by the medium they
1093 are attached to. Two devices can have different id values when
1094 the broadcast packets are received only on one of them.
1095 The default value 0 means that the device is the only interface
1096 to its medium, value of -1 means that medium is not known.
1098 Currently, it is used to change the proxy_arp behavior:
1099 the proxy_arp feature is enabled for packets forwarded between
1100 two devices attached to different media.
1104 proxy_arp for the interface will be enabled if at least one of
1105 conf/{all,interface}/proxy_arp is set to TRUE,
1106 it will be disabled otherwise
1108 proxy_arp_pvlan - BOOLEAN
1109 Private VLAN proxy arp.
1110 Basically allow proxy arp replies back to the same interface
1111 (from which the ARP request/solicitation was received).
1113 This is done to support (ethernet) switch features, like RFC
1114 3069, where the individual ports are NOT allowed to
1115 communicate with each other, but they are allowed to talk to
1116 the upstream router. As described in RFC 3069, it is possible
1117 to allow these hosts to communicate through the upstream
1118 router by proxy_arp'ing. Don't need to be used together with
1121 This technology is known by different names:
1122 In RFC 3069 it is called VLAN Aggregation.
1123 Cisco and Allied Telesyn call it Private VLAN.
1124 Hewlett-Packard call it Source-Port filtering or port-isolation.
1125 Ericsson call it MAC-Forced Forwarding (RFC Draft).
1127 shared_media - BOOLEAN
1128 Send(router) or accept(host) RFC1620 shared media redirects.
1129 Overrides secure_redirects.
1130 shared_media for the interface will be enabled if at least one of
1131 conf/{all,interface}/shared_media is set to TRUE,
1132 it will be disabled otherwise
1135 secure_redirects - BOOLEAN
1136 Accept ICMP redirect messages only to gateways listed in the
1137 interface's current gateway list. Even if disabled, RFC1122 redirect
1139 Overridden by shared_media.
1140 secure_redirects for the interface will be enabled if at least one of
1141 conf/{all,interface}/secure_redirects is set to TRUE,
1142 it will be disabled otherwise
1145 send_redirects - BOOLEAN
1146 Send redirects, if router.
1147 send_redirects for the interface will be enabled if at least one of
1148 conf/{all,interface}/send_redirects is set to TRUE,
1149 it will be disabled otherwise
1152 bootp_relay - BOOLEAN
1153 Accept packets with source address 0.b.c.d destined
1154 not to this host as local ones. It is supposed, that
1155 BOOTP relay daemon will catch and forward such packets.
1156 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1159 Not Implemented Yet.
1161 accept_source_route - BOOLEAN
1162 Accept packets with SRR option.
1163 conf/all/accept_source_route must also be set to TRUE to accept packets
1164 with SRR option on the interface
1165 default TRUE (router)
1168 accept_local - BOOLEAN
1169 Accept packets with local source addresses. In combination with
1170 suitable routing, this can be used to direct packets between two
1171 local interfaces over the wire and have them accepted properly.
1174 route_localnet - BOOLEAN
1175 Do not consider loopback addresses as martian source or destination
1176 while routing. This enables the use of 127/8 for local routing purposes.
1180 0 - No source validation.
1181 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1182 Each incoming packet is tested against the FIB and if the interface
1183 is not the best reverse path the packet check will fail.
1184 By default failed packets are discarded.
1185 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1186 Each incoming packet's source address is also tested against the FIB
1187 and if the source address is not reachable via any interface
1188 the packet check will fail.
1190 Current recommended practice in RFC3704 is to enable strict mode
1191 to prevent IP spoofing from DDos attacks. If using asymmetric routing
1192 or other complicated routing, then loose mode is recommended.
1194 The max value from conf/{all,interface}/rp_filter is used
1195 when doing source validation on the {interface}.
1197 Default value is 0. Note that some distributions enable it
1200 arp_filter - BOOLEAN
1201 1 - Allows you to have multiple network interfaces on the same
1202 subnet, and have the ARPs for each interface be answered
1203 based on whether or not the kernel would route a packet from
1204 the ARP'd IP out that interface (therefore you must use source
1205 based routing for this to work). In other words it allows control
1206 of which cards (usually 1) will respond to an arp request.
1208 0 - (default) The kernel can respond to arp requests with addresses
1209 from other interfaces. This may seem wrong but it usually makes
1210 sense, because it increases the chance of successful communication.
1211 IP addresses are owned by the complete host on Linux, not by
1212 particular interfaces. Only for more complex setups like load-
1213 balancing, does this behaviour cause problems.
1215 arp_filter for the interface will be enabled if at least one of
1216 conf/{all,interface}/arp_filter is set to TRUE,
1217 it will be disabled otherwise
1219 arp_announce - INTEGER
1220 Define different restriction levels for announcing the local
1221 source IP address from IP packets in ARP requests sent on
1223 0 - (default) Use any local address, configured on any interface
1224 1 - Try to avoid local addresses that are not in the target's
1225 subnet for this interface. This mode is useful when target
1226 hosts reachable via this interface require the source IP
1227 address in ARP requests to be part of their logical network
1228 configured on the receiving interface. When we generate the
1229 request we will check all our subnets that include the
1230 target IP and will preserve the source address if it is from
1231 such subnet. If there is no such subnet we select source
1232 address according to the rules for level 2.
1233 2 - Always use the best local address for this target.
1234 In this mode we ignore the source address in the IP packet
1235 and try to select local address that we prefer for talks with
1236 the target host. Such local address is selected by looking
1237 for primary IP addresses on all our subnets on the outgoing
1238 interface that include the target IP address. If no suitable
1239 local address is found we select the first local address
1240 we have on the outgoing interface or on all other interfaces,
1241 with the hope we will receive reply for our request and
1242 even sometimes no matter the source IP address we announce.
1244 The max value from conf/{all,interface}/arp_announce is used.
1246 Increasing the restriction level gives more chance for
1247 receiving answer from the resolved target while decreasing
1248 the level announces more valid sender's information.
1250 arp_ignore - INTEGER
1251 Define different modes for sending replies in response to
1252 received ARP requests that resolve local target IP addresses:
1253 0 - (default): reply for any local target IP address, configured
1255 1 - reply only if the target IP address is local address
1256 configured on the incoming interface
1257 2 - reply only if the target IP address is local address
1258 configured on the incoming interface and both with the
1259 sender's IP address are part from same subnet on this interface
1260 3 - do not reply for local addresses configured with scope host,
1261 only resolutions for global and link addresses are replied
1263 8 - do not reply for all local addresses
1265 The max value from conf/{all,interface}/arp_ignore is used
1266 when ARP request is received on the {interface}
1268 arp_notify - BOOLEAN
1269 Define mode for notification of address and device changes.
1270 0 - (default): do nothing
1271 1 - Generate gratuitous arp requests when device is brought up
1272 or hardware address changes.
1274 arp_accept - BOOLEAN
1275 Define behavior for gratuitous ARP frames who's IP is not
1276 already present in the ARP table:
1277 0 - don't create new entries in the ARP table
1278 1 - create new entries in the ARP table
1280 Both replies and requests type gratuitous arp will trigger the
1281 ARP table to be updated, if this setting is on.
1283 If the ARP table already contains the IP address of the
1284 gratuitous arp frame, the arp table will be updated regardless
1285 if this setting is on or off.
1287 mcast_solicit - INTEGER
1288 The maximum number of multicast probes in INCOMPLETE state,
1289 when the associated hardware address is unknown. Defaults
1292 ucast_solicit - INTEGER
1293 The maximum number of unicast probes in PROBE state, when
1294 the hardware address is being reconfirmed. Defaults to 3.
1296 app_solicit - INTEGER
1297 The maximum number of probes to send to the user space ARP daemon
1298 via netlink before dropping back to multicast probes (see
1299 mcast_resolicit). Defaults to 0.
1301 mcast_resolicit - INTEGER
1302 The maximum number of multicast probes after unicast and
1303 app probes in PROBE state. Defaults to 0.
1305 disable_policy - BOOLEAN
1306 Disable IPSEC policy (SPD) for this interface
1308 disable_xfrm - BOOLEAN
1309 Disable IPSEC encryption on this interface, whatever the policy
1311 igmpv2_unsolicited_report_interval - INTEGER
1312 The interval in milliseconds in which the next unsolicited
1313 IGMPv1 or IGMPv2 report retransmit will take place.
1314 Default: 10000 (10 seconds)
1316 igmpv3_unsolicited_report_interval - INTEGER
1317 The interval in milliseconds in which the next unsolicited
1318 IGMPv3 report retransmit will take place.
1319 Default: 1000 (1 seconds)
1321 promote_secondaries - BOOLEAN
1322 When a primary IP address is removed from this interface
1323 promote a corresponding secondary IP address instead of
1324 removing all the corresponding secondary IP addresses.
1326 drop_unicast_in_l2_multicast - BOOLEAN
1327 Drop any unicast IP packets that are received in link-layer
1328 multicast (or broadcast) frames.
1329 This behavior (for multicast) is actually a SHOULD in RFC
1330 1122, but is disabled by default for compatibility reasons.
1333 drop_gratuitous_arp - BOOLEAN
1334 Drop all gratuitous ARP frames, for example if there's a known
1335 good ARP proxy on the network and such frames need not be used
1336 (or in the case of 802.11, must not be used to prevent attacks.)
1341 Allows you to write a number, which can be used as required.
1344 xfrm4_gc_thresh - INTEGER
1345 (Obsolete since linux-4.14)
1346 The threshold at which we will start garbage collecting for IPv4
1347 destination cache entries. At twice this value the system will
1348 refuse new allocations.
1350 igmp_link_local_mcast_reports - BOOLEAN
1351 Enable IGMP reports for link local multicast groups in the
1356 kuznet@ms2.inr.ac.ru
1362 delon.nicolas@wanadoo.fr
1367 /proc/sys/net/ipv6/* Variables:
1369 IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
1370 apply to IPv6 [XXX?].
1372 bindv6only - BOOLEAN
1373 Default value for IPV6_V6ONLY socket option,
1374 which restricts use of the IPv6 socket to IPv6 communication
1376 TRUE: disable IPv4-mapped address feature
1377 FALSE: enable IPv4-mapped address feature
1379 Default: FALSE (as specified in RFC3493)
1381 flowlabel_consistency - BOOLEAN
1382 Protect the consistency (and unicity) of flow label.
1383 You have to disable it to use IPV6_FL_F_REFLECT flag on the
1389 auto_flowlabels - INTEGER
1390 Automatically generate flow labels based on a flow hash of the
1391 packet. This allows intermediate devices, such as routers, to
1392 identify packet flows for mechanisms like Equal Cost Multipath
1393 Routing (see RFC 6438).
1394 0: automatic flow labels are completely disabled
1395 1: automatic flow labels are enabled by default, they can be
1396 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
1398 2: automatic flow labels are allowed, they may be enabled on a
1399 per socket basis using the IPV6_AUTOFLOWLABEL socket option
1400 3: automatic flow labels are enabled and enforced, they cannot
1401 be disabled by the socket option
1404 flowlabel_state_ranges - BOOLEAN
1405 Split the flow label number space into two ranges. 0-0x7FFFF is
1406 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
1407 is reserved for stateless flow labels as described in RFC6437.
1412 flowlabel_reflect - BOOLEAN
1413 Automatically reflect the flow label. Needed for Path MTU
1414 Discovery to work with Equal Cost Multipath Routing in anycast
1415 environments. See RFC 7690 and:
1416 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
1421 fib_multipath_hash_policy - INTEGER
1422 Controls which hash policy to use for multipath routes.
1423 Default: 0 (Layer 3)
1425 0 - Layer 3 (source and destination addresses plus flow label)
1426 1 - Layer 4 (standard 5-tuple)
1428 anycast_src_echo_reply - BOOLEAN
1429 Controls the use of anycast addresses as source addresses for ICMPv6
1435 idgen_delay - INTEGER
1436 Controls the delay in seconds after which time to retry
1437 privacy stable address generation if a DAD conflict is
1439 Default: 1 (as specified in RFC7217)
1441 idgen_retries - INTEGER
1442 Controls the number of retries to generate a stable privacy
1443 address if a DAD conflict is detected.
1444 Default: 3 (as specified in RFC7217)
1447 Controls the MLD query robustness variable (see RFC3810 9.1).
1448 Default: 2 (as specified by RFC3810 9.1)
1449 Minimum: 1 (as specified by RFC6636 4.5)
1451 max_dst_opts_number - INTEGER
1452 Maximum number of non-padding TLVs allowed in a Destination
1453 options extension header. If this value is less than zero
1454 then unknown options are disallowed and the number of known
1455 TLVs allowed is the absolute value of this number.
1458 max_hbh_opts_number - INTEGER
1459 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
1460 options extension header. If this value is less than zero
1461 then unknown options are disallowed and the number of known
1462 TLVs allowed is the absolute value of this number.
1465 max_dst_opts_length - INTEGER
1466 Maximum length allowed for a Destination options extension
1468 Default: INT_MAX (unlimited)
1470 max_hbh_length - INTEGER
1471 Maximum length allowed for a Hop-by-Hop options extension
1473 Default: INT_MAX (unlimited)
1475 skip_notify_on_dev_down - BOOLEAN
1476 Controls whether an RTM_DELROUTE message is generated for routes
1477 removed when a device is taken down or deleted. IPv4 does not
1478 generate this message; IPv6 does by default. Setting this sysctl
1479 to true skips the message, making IPv4 and IPv6 on par in relying
1480 on userspace caches to track link events and evict routes.
1481 Default: false (generate message)
1485 ip6frag_high_thresh - INTEGER
1486 Maximum memory used to reassemble IPv6 fragments. When
1487 ip6frag_high_thresh bytes of memory is allocated for this purpose,
1488 the fragment handler will toss packets until ip6frag_low_thresh
1491 ip6frag_low_thresh - INTEGER
1492 See ip6frag_high_thresh
1494 ip6frag_time - INTEGER
1495 Time in seconds to keep an IPv6 fragment in memory.
1497 IPv6 Segment Routing:
1499 seg6_flowlabel - INTEGER
1500 Controls the behaviour of computing the flowlabel of outer
1501 IPv6 header in case of SR T.encaps
1503 -1 set flowlabel to zero.
1504 0 copy flowlabel from Inner packet in case of Inner IPv6
1505 (Set flowlabel to 0 in case IPv4/L2)
1506 1 Compute the flowlabel using seg6_make_flowlabel()
1511 Change the interface-specific default settings.
1515 Change all the interface-specific settings.
1517 [XXX: Other special features than forwarding?]
1519 conf/all/forwarding - BOOLEAN
1520 Enable global IPv6 forwarding between all interfaces.
1522 IPv4 and IPv6 work differently here; e.g. netfilter must be used
1523 to control which interfaces may forward packets and which not.
1525 This also sets all interfaces' Host/Router setting
1526 'forwarding' to the specified value. See below for details.
1528 This referred to as global forwarding.
1533 fwmark_reflect - BOOLEAN
1534 Controls the fwmark of kernel-generated IPv6 reply packets that are not
1535 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
1536 If unset, these packets have a fwmark of zero. If set, they have the
1537 fwmark of the packet they are replying to.
1541 Change special settings per interface.
1543 The functional behaviour for certain settings is different
1544 depending on whether local forwarding is enabled or not.
1547 Accept Router Advertisements; autoconfigure using them.
1549 It also determines whether or not to transmit Router
1550 Solicitations. If and only if the functional setting is to
1551 accept Router Advertisements, Router Solicitations will be
1554 Possible values are:
1555 0 Do not accept Router Advertisements.
1556 1 Accept Router Advertisements if forwarding is disabled.
1557 2 Overrule forwarding behaviour. Accept Router Advertisements
1558 even if forwarding is enabled.
1560 Functional default: enabled if local forwarding is disabled.
1561 disabled if local forwarding is enabled.
1563 accept_ra_defrtr - BOOLEAN
1564 Learn default router in Router Advertisement.
1566 Functional default: enabled if accept_ra is enabled.
1567 disabled if accept_ra is disabled.
1569 accept_ra_from_local - BOOLEAN
1570 Accept RA with source-address that is found on local machine
1571 if the RA is otherwise proper and able to be accepted.
1572 Default is to NOT accept these as it may be an un-intended
1576 enabled if accept_ra_from_local is enabled
1577 on a specific interface.
1578 disabled if accept_ra_from_local is disabled
1579 on a specific interface.
1581 accept_ra_min_hop_limit - INTEGER
1582 Minimum hop limit Information in Router Advertisement.
1584 Hop limit Information in Router Advertisement less than this
1585 variable shall be ignored.
1589 accept_ra_pinfo - BOOLEAN
1590 Learn Prefix Information in Router Advertisement.
1592 Functional default: enabled if accept_ra is enabled.
1593 disabled if accept_ra is disabled.
1595 accept_ra_rt_info_min_plen - INTEGER
1596 Minimum prefix length of Route Information in RA.
1598 Route Information w/ prefix smaller than this variable shall
1601 Functional default: 0 if accept_ra_rtr_pref is enabled.
1602 -1 if accept_ra_rtr_pref is disabled.
1604 accept_ra_rt_info_max_plen - INTEGER
1605 Maximum prefix length of Route Information in RA.
1607 Route Information w/ prefix larger than this variable shall
1610 Functional default: 0 if accept_ra_rtr_pref is enabled.
1611 -1 if accept_ra_rtr_pref is disabled.
1613 accept_ra_rtr_pref - BOOLEAN
1614 Accept Router Preference in RA.
1616 Functional default: enabled if accept_ra is enabled.
1617 disabled if accept_ra is disabled.
1619 accept_ra_mtu - BOOLEAN
1620 Apply the MTU value specified in RA option 5 (RFC4861). If
1621 disabled, the MTU specified in the RA will be ignored.
1623 Functional default: enabled if accept_ra is enabled.
1624 disabled if accept_ra is disabled.
1626 accept_redirects - BOOLEAN
1629 Functional default: enabled if local forwarding is disabled.
1630 disabled if local forwarding is enabled.
1632 accept_source_route - INTEGER
1633 Accept source routing (routing extension header).
1635 >= 0: Accept only routing header type 2.
1636 < 0: Do not accept routing header.
1641 Autoconfigure addresses using Prefix Information in Router
1644 Functional default: enabled if accept_ra_pinfo is enabled.
1645 disabled if accept_ra_pinfo is disabled.
1647 dad_transmits - INTEGER
1648 The amount of Duplicate Address Detection probes to send.
1651 forwarding - INTEGER
1652 Configure interface-specific Host/Router behaviour.
1654 Note: It is recommended to have the same setting on all
1655 interfaces; mixed router/host scenarios are rather uncommon.
1657 Possible values are:
1658 0 Forwarding disabled
1659 1 Forwarding enabled
1663 By default, Host behaviour is assumed. This means:
1665 1. IsRouter flag is not set in Neighbour Advertisements.
1666 2. If accept_ra is TRUE (default), transmit Router
1668 3. If accept_ra is TRUE (default), accept Router
1669 Advertisements (and do autoconfiguration).
1670 4. If accept_redirects is TRUE (default), accept Redirects.
1674 If local forwarding is enabled, Router behaviour is assumed.
1675 This means exactly the reverse from the above:
1677 1. IsRouter flag is set in Neighbour Advertisements.
1678 2. Router Solicitations are not sent unless accept_ra is 2.
1679 3. Router Advertisements are ignored unless accept_ra is 2.
1680 4. Redirects are ignored.
1682 Default: 0 (disabled) if global forwarding is disabled (default),
1683 otherwise 1 (enabled).
1686 Default Hop Limit to set.
1690 Default Maximum Transfer Unit
1691 Default: 1280 (IPv6 required minimum)
1693 ip_nonlocal_bind - BOOLEAN
1694 If set, allows processes to bind() to non-local IPv6 addresses,
1695 which can be quite useful - but may break some applications.
1698 router_probe_interval - INTEGER
1699 Minimum interval (in seconds) between Router Probing described
1704 router_solicitation_delay - INTEGER
1705 Number of seconds to wait after interface is brought up
1706 before sending Router Solicitations.
1709 router_solicitation_interval - INTEGER
1710 Number of seconds to wait between Router Solicitations.
1713 router_solicitations - INTEGER
1714 Number of Router Solicitations to send until assuming no
1715 routers are present.
1718 use_oif_addrs_only - BOOLEAN
1719 When enabled, the candidate source addresses for destinations
1720 routed via this interface are restricted to the set of addresses
1721 configured on this interface (vis. RFC 6724, section 4).
1725 use_tempaddr - INTEGER
1726 Preference for Privacy Extensions (RFC3041).
1727 <= 0 : disable Privacy Extensions
1728 == 1 : enable Privacy Extensions, but prefer public
1729 addresses over temporary addresses.
1730 > 1 : enable Privacy Extensions and prefer temporary
1731 addresses over public addresses.
1732 Default: 0 (for most devices)
1733 -1 (for point-to-point devices and loopback devices)
1735 temp_valid_lft - INTEGER
1736 valid lifetime (in seconds) for temporary addresses.
1737 Default: 604800 (7 days)
1739 temp_prefered_lft - INTEGER
1740 Preferred lifetime (in seconds) for temporary addresses.
1741 Default: 86400 (1 day)
1743 keep_addr_on_down - INTEGER
1744 Keep all IPv6 addresses on an interface down event. If set static
1745 global addresses with no expiration time are not flushed.
1750 Default: 0 (addresses are removed)
1752 max_desync_factor - INTEGER
1753 Maximum value for DESYNC_FACTOR, which is a random value
1754 that ensures that clients don't synchronize with each
1755 other and generate new addresses at exactly the same time.
1756 value is in seconds.
1759 regen_max_retry - INTEGER
1760 Number of attempts before give up attempting to generate
1761 valid temporary addresses.
1764 max_addresses - INTEGER
1765 Maximum number of autoconfigured addresses per interface. Setting
1766 to zero disables the limitation. It is not recommended to set this
1767 value too large (or to zero) because it would be an easy way to
1768 crash the kernel by allowing too many addresses to be created.
1771 disable_ipv6 - BOOLEAN
1772 Disable IPv6 operation. If accept_dad is set to 2, this value
1773 will be dynamically set to TRUE if DAD fails for the link-local
1775 Default: FALSE (enable IPv6 operation)
1777 When this value is changed from 1 to 0 (IPv6 is being enabled),
1778 it will dynamically create a link-local address on the given
1779 interface and start Duplicate Address Detection, if necessary.
1781 When this value is changed from 0 to 1 (IPv6 is being disabled),
1782 it will dynamically delete all addresses and routes on the given
1783 interface. From now on it will not possible to add addresses/routes
1784 to the selected interface.
1786 accept_dad - INTEGER
1787 Whether to accept DAD (Duplicate Address Detection).
1789 1: Enable DAD (default)
1790 2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1791 link-local address has been found.
1793 DAD operation and mode on a given interface will be selected according
1794 to the maximum value of conf/{all,interface}/accept_dad.
1796 force_tllao - BOOLEAN
1797 Enable sending the target link-layer address option even when
1798 responding to a unicast neighbor solicitation.
1801 Quoting from RFC 2461, section 4.4, Target link-layer address:
1803 "The option MUST be included for multicast solicitations in order to
1804 avoid infinite Neighbor Solicitation "recursion" when the peer node
1805 does not have a cache entry to return a Neighbor Advertisements
1806 message. When responding to unicast solicitations, the option can be
1807 omitted since the sender of the solicitation has the correct link-
1808 layer address; otherwise it would not have be able to send the unicast
1809 solicitation in the first place. However, including the link-layer
1810 address in this case adds little overhead and eliminates a potential
1811 race condition where the sender deletes the cached link-layer address
1812 prior to receiving a response to a previous solicitation."
1814 ndisc_notify - BOOLEAN
1815 Define mode for notification of address and device changes.
1816 0 - (default): do nothing
1817 1 - Generate unsolicited neighbour advertisements when device is brought
1818 up or hardware address changes.
1820 ndisc_tclass - INTEGER
1821 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
1822 Discovery (Router Solicitation, Router Advertisement, Neighbor
1823 Solicitation, Neighbor Advertisement, Redirect) messages.
1824 These 8 bits can be interpreted as 6 high order bits holding the DSCP
1825 value and 2 low order bits representing ECN (which you probably want
1829 mldv1_unsolicited_report_interval - INTEGER
1830 The interval in milliseconds in which the next unsolicited
1831 MLDv1 report retransmit will take place.
1832 Default: 10000 (10 seconds)
1834 mldv2_unsolicited_report_interval - INTEGER
1835 The interval in milliseconds in which the next unsolicited
1836 MLDv2 report retransmit will take place.
1837 Default: 1000 (1 second)
1839 force_mld_version - INTEGER
1840 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1841 1 - Enforce to use MLD version 1
1842 2 - Enforce to use MLD version 2
1844 suppress_frag_ndisc - INTEGER
1845 Control RFC 6980 (Security Implications of IPv6 Fragmentation
1846 with IPv6 Neighbor Discovery) behavior:
1847 1 - (default) discard fragmented neighbor discovery packets
1848 0 - allow fragmented neighbor discovery packets
1850 optimistic_dad - BOOLEAN
1851 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
1852 0: disabled (default)
1855 Optimistic Duplicate Address Detection for the interface will be enabled
1856 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
1857 it will be disabled otherwise.
1859 use_optimistic - BOOLEAN
1860 If enabled, do not classify optimistic addresses as deprecated during
1861 source address selection. Preferred addresses will still be chosen
1862 before optimistic addresses, subject to other ranking in the source
1863 address selection algorithm.
1864 0: disabled (default)
1867 This will be enabled if at least one of
1868 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
1870 stable_secret - IPv6 address
1871 This IPv6 address will be used as a secret to generate IPv6
1872 addresses for link-local addresses and autoconfigured
1873 ones. All addresses generated after setting this secret will
1874 be stable privacy ones by default. This can be changed via the
1875 addrgenmode ip-link. conf/default/stable_secret is used as the
1876 secret for the namespace, the interface specific ones can
1877 overwrite that. Writes to conf/all/stable_secret are refused.
1879 It is recommended to generate this secret during installation
1880 of a system and keep it stable after that.
1882 By default the stable secret is unset.
1884 addr_gen_mode - INTEGER
1885 Defines how link-local and autoconf addresses are generated.
1887 0: generate address based on EUI64 (default)
1888 1: do no generate a link-local address, use EUI64 for addresses generated
1890 2: generate stable privacy addresses, using the secret from
1891 stable_secret (RFC7217)
1892 3: generate stable privacy addresses, using a random secret if unset
1894 drop_unicast_in_l2_multicast - BOOLEAN
1895 Drop any unicast IPv6 packets that are received in link-layer
1896 multicast (or broadcast) frames.
1898 By default this is turned off.
1900 drop_unsolicited_na - BOOLEAN
1901 Drop all unsolicited neighbor advertisements, for example if there's
1902 a known good NA proxy on the network and such frames need not be used
1903 (or in the case of 802.11, must not be used to prevent attacks.)
1905 By default this is turned off.
1907 enhanced_dad - BOOLEAN
1908 Include a nonce option in the IPv6 neighbor solicitation messages used for
1909 duplicate address detection per RFC7527. A received DAD NS will only signal
1910 a duplicate address if the nonce is different. This avoids any false
1911 detection of duplicates due to loopback of the NS messages that we send.
1912 The nonce option will be sent on an interface unless both of
1913 conf/{all,interface}/enhanced_dad are set to FALSE.
1918 Limit the maximal rates for sending ICMPv6 messages.
1919 0 to disable any limiting,
1920 otherwise the minimal space between responses in milliseconds.
1923 ratemask - list of comma separated ranges
1924 For ICMPv6 message types matching the ranges in the ratemask, limit
1925 the sending of the message according to ratelimit parameter.
1927 The format used for both input and output is a comma separated
1928 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
1929 129). Writing to the file will clear all previous ranges of ICMPv6
1930 message types and update the current list with the input.
1932 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
1933 for numerical values of ICMPv6 message types, e.g. echo request is 128
1934 and echo reply is 129.
1936 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
1938 echo_ignore_all - BOOLEAN
1939 If set non-zero, then the kernel will ignore all ICMP ECHO
1940 requests sent to it over the IPv6 protocol.
1943 echo_ignore_multicast - BOOLEAN
1944 If set non-zero, then the kernel will ignore all ICMP ECHO
1945 requests sent to it over the IPv6 protocol via multicast.
1948 echo_ignore_anycast - BOOLEAN
1949 If set non-zero, then the kernel will ignore all ICMP ECHO
1950 requests sent to it over the IPv6 protocol destined to anycast address.
1953 xfrm6_gc_thresh - INTEGER
1954 (Obsolete since linux-4.14)
1955 The threshold at which we will start garbage collecting for IPv6
1956 destination cache entries. At twice this value the system will
1957 refuse new allocations.
1961 Pekka Savola <pekkas@netcore.fi>
1962 YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1965 /proc/sys/net/bridge/* Variables:
1967 bridge-nf-call-arptables - BOOLEAN
1968 1 : pass bridged ARP traffic to arptables' FORWARD chain.
1972 bridge-nf-call-iptables - BOOLEAN
1973 1 : pass bridged IPv4 traffic to iptables' chains.
1977 bridge-nf-call-ip6tables - BOOLEAN
1978 1 : pass bridged IPv6 traffic to ip6tables' chains.
1982 bridge-nf-filter-vlan-tagged - BOOLEAN
1983 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1987 bridge-nf-filter-pppoe-tagged - BOOLEAN
1988 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1992 bridge-nf-pass-vlan-input-dev - BOOLEAN
1993 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1994 interface on the bridge and set the netfilter input device to the vlan.
1995 This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1996 target work with vlan-on-top-of-bridge interfaces. When no matching
1997 vlan interface is found, or this switch is off, the input device is
1998 set to the bridge interface.
1999 0: disable bridge netfilter vlan interface lookup.
2002 proc/sys/net/sctp/* Variables:
2004 addip_enable - BOOLEAN
2005 Enable or disable extension of Dynamic Address Reconfiguration
2006 (ADD-IP) functionality specified in RFC5061. This extension provides
2007 the ability to dynamically add and remove new addresses for the SCTP
2010 1: Enable extension.
2012 0: Disable extension.
2017 Enable or disable pf (pf is short for potentially failed) state. A value
2018 of pf_retrans > path_max_retrans also disables pf state. That is, one of
2019 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2020 Since pf_retrans and path_max_retrans can be changed by userspace
2021 application, sometimes user expects to disable pf state by the value of
2022 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2023 or path_max_retrans is changed by the user application, this pf state is
2024 enabled. As such, it is necessary to add this to dynamically enable
2025 and disable pf state. See:
2026 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2035 addip_noauth_enable - BOOLEAN
2036 Dynamic Address Reconfiguration (ADD-IP) requires the use of
2037 authentication to protect the operations of adding or removing new
2038 addresses. This requirement is mandated so that unauthorized hosts
2039 would not be able to hijack associations. However, older
2040 implementations may not have implemented this requirement while
2041 allowing the ADD-IP extension. For reasons of interoperability,
2042 we provide this variable to control the enforcement of the
2043 authentication requirement.
2045 1: Allow ADD-IP extension to be used without authentication. This
2046 should only be set in a closed environment for interoperability
2047 with older implementations.
2049 0: Enforce the authentication requirement
2053 auth_enable - BOOLEAN
2054 Enable or disable Authenticated Chunks extension. This extension
2055 provides the ability to send and receive authenticated chunks and is
2056 required for secure operation of Dynamic Address Reconfiguration
2059 1: Enable this extension.
2060 0: Disable this extension.
2064 prsctp_enable - BOOLEAN
2065 Enable or disable the Partial Reliability extension (RFC3758) which
2066 is used to notify peers that a given DATA should no longer be expected.
2074 The limit of the number of new packets that can be initially sent. It
2075 controls how bursty the generated traffic can be.
2079 association_max_retrans - INTEGER
2080 Set the maximum number for retransmissions that an association can
2081 attempt deciding that the remote end is unreachable. If this value
2082 is exceeded, the association is terminated.
2086 max_init_retransmits - INTEGER
2087 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
2088 that an association will attempt before declaring the destination
2089 unreachable and terminating.
2093 path_max_retrans - INTEGER
2094 The maximum number of retransmissions that will be attempted on a given
2095 path. Once this threshold is exceeded, the path is considered
2096 unreachable, and new traffic will use a different path when the
2097 association is multihomed.
2101 pf_retrans - INTEGER
2102 The number of retransmissions that will be attempted on a given path
2103 before traffic is redirected to an alternate transport (should one
2104 exist). Note this is distinct from path_max_retrans, as a path that
2105 passes the pf_retrans threshold can still be used. Its only
2106 deprioritized when a transmission path is selected by the stack. This
2107 setting is primarily used to enable fast failover mechanisms without
2108 having to reduce path_max_retrans to a very low value. See:
2109 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
2110 for details. Note also that a value of pf_retrans > path_max_retrans
2111 disables this feature. Since both pf_retrans and path_max_retrans can
2112 be changed by userspace application, a variable pf_enable is used to
2117 rto_initial - INTEGER
2118 The initial round trip timeout value in milliseconds that will be used
2119 in calculating round trip times. This is the initial time interval
2120 for retransmissions.
2125 The maximum value (in milliseconds) of the round trip timeout. This
2126 is the largest time interval that can elapse between retransmissions.
2131 The minimum value (in milliseconds) of the round trip timeout. This
2132 is the smallest time interval the can elapse between retransmissions.
2136 hb_interval - INTEGER
2137 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
2138 are sent at the specified interval on idle paths to probe the state of
2139 a given path between 2 associations.
2143 sack_timeout - INTEGER
2144 The amount of time (in milliseconds) that the implementation will wait
2149 valid_cookie_life - INTEGER
2150 The default lifetime of the SCTP cookie (in milliseconds). The cookie
2151 is used during association establishment.
2155 cookie_preserve_enable - BOOLEAN
2156 Enable or disable the ability to extend the lifetime of the SCTP cookie
2157 that is used during the establishment phase of SCTP association
2159 1: Enable cookie lifetime extension.
2164 cookie_hmac_alg - STRING
2165 Select the hmac algorithm used when generating the cookie value sent by
2166 a listening sctp socket to a connecting client in the INIT-ACK chunk.
2171 Ability to assign md5 or sha1 as the selected alg is predicated on the
2172 configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
2173 CONFIG_CRYPTO_SHA1).
2175 Default: Dependent on configuration. MD5 if available, else SHA1 if
2176 available, else none.
2178 rcvbuf_policy - INTEGER
2179 Determines if the receive buffer is attributed to the socket or to
2180 association. SCTP supports the capability to create multiple
2181 associations on a single socket. When using this capability, it is
2182 possible that a single stalled association that's buffering a lot
2183 of data may block other associations from delivering their data by
2184 consuming all of the receive buffer space. To work around this,
2185 the rcvbuf_policy could be set to attribute the receiver buffer space
2186 to each association instead of the socket. This prevents the described
2189 1: rcvbuf space is per association
2190 0: rcvbuf space is per socket
2194 sndbuf_policy - INTEGER
2195 Similar to rcvbuf_policy above, this applies to send buffer space.
2197 1: Send buffer is tracked per association
2198 0: Send buffer is tracked per socket.
2202 sctp_mem - vector of 3 INTEGERs: min, pressure, max
2203 Number of pages allowed for queueing by all SCTP sockets.
2205 min: Below this number of pages SCTP is not bothered about its
2206 memory appetite. When amount of memory allocated by SCTP exceeds
2207 this number, SCTP starts to moderate memory usage.
2209 pressure: This value was introduced to follow format of tcp_mem.
2211 max: Number of pages allowed for queueing by all SCTP sockets.
2213 Default is calculated at boot time from amount of available memory.
2215 sctp_rmem - vector of 3 INTEGERs: min, default, max
2216 Only the first value ("min") is used, "default" and "max" are
2219 min: Minimal size of receive buffer used by SCTP socket.
2220 It is guaranteed to each SCTP socket (but not association) even
2221 under moderate memory pressure.
2225 sctp_wmem - vector of 3 INTEGERs: min, default, max
2226 Currently this tunable has no effect.
2228 addr_scope_policy - INTEGER
2229 Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
2231 0 - Disable IPv4 address scoping
2232 1 - Enable IPv4 address scoping
2233 2 - Follow draft but allow IPv4 private addresses
2234 3 - Follow draft but allow IPv4 link local addresses
2239 /proc/sys/net/core/*
2240 Please see: Documentation/sysctl/net.txt for descriptions of these entries.
2243 /proc/sys/net/unix/*
2244 max_dgram_qlen - INTEGER
2245 The maximum length of dgram socket receive queue