Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1.. SPDX-License-Identifier: GPL-2.0
2
3=========
4IP Sysctl
5=========
6
7/proc/sys/net/ipv4/* Variables
8==============================
9
10ip_forward - BOOLEAN
11 Forward Packets between interfaces.
12
13 This variable is special, its change resets all configuration
14 parameters to their default state (RFC1122 for hosts, RFC1812
15 for routers)
16
17 Possible values:
18
19 - 0 (disabled)
20 - 1 (enabled)
21
22 Default: 0 (disabled)
23
24ip_default_ttl - INTEGER
25 Default value of TTL field (Time To Live) for outgoing (but not
26 forwarded) IP packets. Should be between 1 and 255 inclusive.
27 Default: 64 (as recommended by RFC1700)
28
29ip_no_pmtu_disc - INTEGER
30 Disable Path MTU Discovery. If enabled in mode 1 and a
31 fragmentation-required ICMP is received, the PMTU to this
32 destination will be set to the smallest of the old MTU to
33 this destination and min_pmtu (see below). You will need
34 to raise min_pmtu to the smallest interface MTU on your system
35 manually if you want to avoid locally generated fragments.
36
37 In mode 2 incoming Path MTU Discovery messages will be
38 discarded. Outgoing frames are handled the same as in mode 1,
39 implicitly setting IP_PMTUDISC_DONT on every created socket.
40
41 Mode 3 is a hardened pmtu discover mode. The kernel will only
42 accept fragmentation-needed errors if the underlying protocol
43 can verify them besides a plain socket lookup. Current
44 protocols for which pmtu events will be honored are TCP and
45 SCTP as they verify e.g. the sequence number or the
46 association. This mode should not be enabled globally but is
47 only intended to secure e.g. name servers in namespaces where
48 TCP path mtu must still work but path MTU information of other
49 protocols should be discarded. If enabled globally this mode
50 could break other protocols.
51
52 Possible values: 0-3
53
54 Default: FALSE
55
56min_pmtu - INTEGER
57 default 552 - minimum Path MTU. Unless this is changed manually,
58 each cached pmtu will never be lower than this setting.
59
60ip_forward_use_pmtu - BOOLEAN
61 By default we don't trust protocol path MTUs while forwarding
62 because they could be easily forged and can lead to unwanted
63 fragmentation by the router.
64 You only need to enable this if you have user-space software
65 which tries to discover path mtus by itself and depends on the
66 kernel honoring this information. This is normally not the
67 case.
68
69 Possible values:
70
71 - 0 (disabled)
72 - 1 (enabled)
73
74 Default: 0 (disabled)
75
76fwmark_reflect - BOOLEAN
77 Controls the fwmark of kernel-generated IPv4 reply packets that are not
78 associated with a socket for example, TCP RSTs or ICMP echo replies).
79 If disabled, these packets have a fwmark of zero. If enabled, they have the
80 fwmark of the packet they are replying to.
81
82 Possible values:
83
84 - 0 (disabled)
85 - 1 (enabled)
86
87 Default: 0 (disabled)
88
89fib_multipath_use_neigh - BOOLEAN
90 Use status of existing neighbor entry when determining nexthop for
91 multipath routes. If disabled, neighbor information is not used and
92 packets could be directed to a failed nexthop. Only valid for kernels
93 built with CONFIG_IP_ROUTE_MULTIPATH enabled.
94
95 Possible values:
96
97 - 0 (disabled)
98 - 1 (enabled)
99
100 Default: 0 (disabled)
101
102fib_multipath_hash_policy - INTEGER
103 Controls which hash policy to use for multipath routes. Only valid
104 for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
105
106 Default: 0 (Layer 3)
107
108 Possible values:
109
110 - 0 - Layer 3
111 - 1 - Layer 4
112 - 2 - Layer 3 or inner Layer 3 if present
113 - 3 - Custom multipath hash. Fields used for multipath hash calculation
114 are determined by fib_multipath_hash_fields sysctl
115
116fib_multipath_hash_fields - UNSIGNED INTEGER
117 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
118 fields used for multipath hash calculation are determined by this
119 sysctl.
120
121 This value is a bitmask which enables various fields for multipath hash
122 calculation.
123
124 Possible fields are:
125
126 ====== ============================
127 0x0001 Source IP address
128 0x0002 Destination IP address
129 0x0004 IP protocol
130 0x0008 Unused (Flow Label)
131 0x0010 Source port
132 0x0020 Destination port
133 0x0040 Inner source IP address
134 0x0080 Inner destination IP address
135 0x0100 Inner IP protocol
136 0x0200 Inner Flow Label
137 0x0400 Inner source port
138 0x0800 Inner destination port
139 ====== ============================
140
141 Default: 0x0007 (source IP, destination IP and IP protocol)
142
143fib_multipath_hash_seed - UNSIGNED INTEGER
144 The seed value used when calculating hash for multipath routes. Applies
145 to both IPv4 and IPv6 datapath. Only present for kernels built with
146 CONFIG_IP_ROUTE_MULTIPATH enabled.
147
148 When set to 0, the seed value used for multipath routing defaults to an
149 internal random-generated one.
150
151 The actual hashing algorithm is not specified -- there is no guarantee
152 that a next hop distribution effected by a given seed will keep stable
153 across kernel versions.
154
155 Default: 0 (random)
156
157fib_sync_mem - UNSIGNED INTEGER
158 Amount of dirty memory from fib entries that can be backlogged before
159 synchronize_rcu is forced.
160
161 Default: 512kB Minimum: 64kB Maximum: 64MB
162
163ip_forward_update_priority - INTEGER
164 Whether to update SKB priority from "TOS" field in IPv4 header after it
165 is forwarded. The new SKB priority is mapped from TOS field value
166 according to an rt_tos2priority table (see e.g. man tc-prio).
167
168 Default: 1 (Update priority.)
169
170 Possible values:
171
172 - 0 - Do not update priority.
173 - 1 - Update priority.
174
175route/max_size - INTEGER
176 Maximum number of routes allowed in the kernel. Increase
177 this when using large numbers of interfaces and/or routes.
178
179 From linux kernel 3.6 onwards, this is deprecated for ipv4
180 as route cache is no longer used.
181
182 From linux kernel 6.3 onwards, this is deprecated for ipv6
183 as garbage collection manages cached route entries.
184
185neigh/default/gc_thresh1 - INTEGER
186 Minimum number of entries to keep. Garbage collector will not
187 purge entries if there are fewer than this number.
188
189 Default: 128
190
191neigh/default/gc_thresh2 - INTEGER
192 Threshold when garbage collector becomes more aggressive about
193 purging entries. Entries older than 5 seconds will be cleared
194 when over this number.
195
196 Default: 512
197
198neigh/default/gc_thresh3 - INTEGER
199 Maximum number of non-PERMANENT neighbor entries allowed. Increase
200 this when using large numbers of interfaces and when communicating
201 with large numbers of directly-connected peers.
202
203 Default: 1024
204
205neigh/default/gc_interval - INTEGER
206 Specifies how often the garbage collector for neighbor entries
207 should run. This value applies to the entire table, not
208 individual entries. Unused since kernel v2.6.8.
209
210 Default: 30 seconds
211
212neigh/default/gc_stale_time - INTEGER
213 Determines how long a neighbor entry can remain unused before it is
214 considered stale and eligible for garbage collection. Entries that have
215 not been used for longer than this time will be removed by the garbage
216 collector, unless they have active references, are marked as PERMANENT,
217 or carry the NTF_EXT_LEARNED or NTF_EXT_VALIDATED flag. Stale entries
218 are only removed by the periodic GC when there are at least gc_thresh1
219 neighbors in the table.
220
221 Default: 60 seconds
222
223neigh/default/unres_qlen_bytes - INTEGER
224 The maximum number of bytes which may be used by packets
225 queued for each unresolved address by other network layers.
226 (added in linux 3.3)
227
228 Setting negative value is meaningless and will return error.
229
230 Default: SK_WMEM_DEFAULT, (same as net.core.wmem_default).
231
232 Exact value depends on architecture and kernel options,
233 but should be enough to allow queuing 256 packets
234 of medium size.
235
236neigh/default/unres_qlen - INTEGER
237 The maximum number of packets which may be queued for each
238 unresolved address by other network layers.
239
240 (deprecated in linux 3.3) : use unres_qlen_bytes instead.
241
242 Prior to linux 3.3, the default value is 3 which may cause
243 unexpected packet loss. The current default value is calculated
244 according to default value of unres_qlen_bytes and true size of
245 packet.
246
247 Default: 101
248
249neigh/default/interval_probe_time_ms - INTEGER
250 The probe interval for neighbor entries with NTF_MANAGED flag,
251 the min value is 1.
252
253 Default: 5000
254
255mtu_expires - INTEGER
256 Time, in seconds, that cached PMTU information is kept.
257
258min_adv_mss - INTEGER
259 The advertised MSS depends on the first hop route MTU, but will
260 never be lower than this setting.
261
262fib_notify_on_flag_change - INTEGER
263 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
264 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
265
266 After installing a route to the kernel, user space receives an
267 acknowledgment, which means the route was installed in the kernel,
268 but not necessarily in hardware.
269 It is also possible for a route already installed in hardware to change
270 its action and therefore its flags. For example, a host route that is
271 trapping packets can be "promoted" to perform decapsulation following
272 the installation of an IPinIP/VXLAN tunnel.
273 The notifications will indicate to user-space the state of the route.
274
275 Default: 0 (Do not emit notifications.)
276
277 Possible values:
278
279 - 0 - Do not emit notifications.
280 - 1 - Emit notifications.
281 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
282
283IP Fragmentation:
284
285ipfrag_high_thresh - LONG INTEGER
286 Maximum memory used to reassemble IP fragments.
287
288ipfrag_low_thresh - LONG INTEGER
289 (Obsolete since linux-4.17)
290 Maximum memory used to reassemble IP fragments before the kernel
291 begins to remove incomplete fragment queues to free up resources.
292 The kernel still accepts new fragments for defragmentation.
293
294ipfrag_time - INTEGER
295 Time in seconds to keep an IP fragment in memory.
296
297ipfrag_max_dist - INTEGER
298 ipfrag_max_dist is a non-negative integer value which defines the
299 maximum "disorder" which is allowed among fragments which share a
300 common IP source address. Note that reordering of packets is
301 not unusual, but if a large number of fragments arrive from a source
302 IP address while a particular fragment queue remains incomplete, it
303 probably indicates that one or more fragments belonging to that queue
304 have been lost. When ipfrag_max_dist is positive, an additional check
305 is done on fragments before they are added to a reassembly queue - if
306 ipfrag_max_dist (or more) fragments have arrived from a particular IP
307 address between additions to any IP fragment queue using that source
308 address, it's presumed that one or more fragments in the queue are
309 lost. The existing fragment queue will be dropped, and a new one
310 started. An ipfrag_max_dist value of zero disables this check.
311
312 Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
313 result in unnecessarily dropping fragment queues when normal
314 reordering of packets occurs, which could lead to poor application
315 performance. Using a very large value, e.g. 50000, increases the
316 likelihood of incorrectly reassembling IP fragments that originate
317 from different IP datagrams, which could result in data corruption.
318 Default: 64
319
320bc_forwarding - INTEGER
321 bc_forwarding enables the feature described in rfc1812#section-5.3.5.2
322 and rfc2644. It allows the router to forward directed broadcast.
323 To enable this feature, the 'all' entry and the input interface entry
324 should be set to 1.
325 Default: 0
326
327INET peer storage
328=================
329
330inet_peer_threshold - INTEGER
331 The approximate size of the storage. Starting from this threshold
332 entries will be thrown aggressively. This threshold also determines
333 entries' time-to-live and time intervals between garbage collection
334 passes. More entries, less time-to-live, less GC interval.
335
336inet_peer_minttl - INTEGER
337 Minimum time-to-live of entries. Should be enough to cover fragment
338 time-to-live on the reassembling side. This minimum time-to-live is
339 guaranteed if the pool size is less than inet_peer_threshold.
340 Measured in seconds.
341
342inet_peer_maxttl - INTEGER
343 Maximum time-to-live of entries. Unused entries will expire after
344 this period of time if there is no memory pressure on the pool (i.e.
345 when the number of entries in the pool is very small).
346 Measured in seconds.
347
348TCP variables
349=============
350
351somaxconn - INTEGER
352 Limit of socket listen() backlog, known in userspace as SOMAXCONN.
353 Defaults to 4096. (Was 128 before linux-5.4)
354 See also tcp_max_syn_backlog for additional tuning for TCP sockets.
355
356tcp_abort_on_overflow - BOOLEAN
357 If listening service is too slow to accept new connections,
358 reset them. Default state is FALSE. It means that if overflow
359 occurred due to a burst, connection will recover. Enable this
360 option _only_ if you are really sure that listening daemon
361 cannot be tuned to accept connections faster. Enabling this
362 option can harm clients of your server.
363
364tcp_adv_win_scale - INTEGER
365 Obsolete since linux-6.6
366 Count buffering overhead as bytes/2^tcp_adv_win_scale
367 (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
368 if it is <= 0.
369
370 Possible values are [-31, 31], inclusive.
371
372 Default: 1
373
374tcp_allowed_congestion_control - STRING
375 Show/set the congestion control choices available to non-privileged
376 processes. The list is a subset of those listed in
377 tcp_available_congestion_control.
378
379 Default is "reno" and the default setting (tcp_congestion_control).
380
381tcp_app_win - INTEGER
382 Reserve max(window/2^tcp_app_win, mss) of window for application
383 buffer. Value 0 is special, it means that nothing is reserved.
384
385 Possible values are [0, 31], inclusive.
386
387 Default: 31
388
389tcp_autocorking - BOOLEAN
390 Enable TCP auto corking :
391 When applications do consecutive small write()/sendmsg() system calls,
392 we try to coalesce these small writes as much as possible, to lower
393 total amount of sent packets. This is done if at least one prior
394 packet for the flow is waiting in Qdisc queues or device transmit
395 queue. Applications can still use TCP_CORK for optimal behavior
396 when they know how/when to uncork their sockets.
397
398 Possible values:
399
400 - 0 (disabled)
401 - 1 (enabled)
402
403 Default: 1 (enabled)
404
405tcp_available_congestion_control - STRING
406 Shows the available congestion control choices that are registered.
407 More congestion control algorithms may be available as modules,
408 but not loaded.
409
410tcp_base_mss - INTEGER
411 The initial value of search_low to be used by the packetization layer
412 Path MTU discovery (MTU probing). If MTU probing is enabled,
413 this is the initial MSS used by the connection.
414
415tcp_mtu_probe_floor - INTEGER
416 If MTU probing is enabled this caps the minimum MSS used for search_low
417 for the connection.
418
419 Default : 48
420
421tcp_min_snd_mss - INTEGER
422 TCP SYN and SYNACK messages usually advertise an ADVMSS option,
423 as described in RFC 1122 and RFC 6691.
424
425 If this ADVMSS option is smaller than tcp_min_snd_mss,
426 it is silently capped to tcp_min_snd_mss.
427
428 Default : 48 (at least 8 bytes of payload per segment)
429
430tcp_congestion_control - STRING
431 Set the congestion control algorithm to be used for new
432 connections. The algorithm "reno" is always available, but
433 additional choices may be available based on kernel configuration.
434 Default is set as part of kernel configuration.
435 For passive connections, the listener congestion control choice
436 is inherited.
437
438 [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
439
440tcp_dsack - BOOLEAN
441 Allows TCP to send "duplicate" SACKs.
442
443 Possible values:
444
445 - 0 (disabled)
446 - 1 (enabled)
447
448 Default: 1 (enabled)
449
450tcp_early_retrans - INTEGER
451 Tail loss probe (TLP) converts RTOs occurring due to tail
452 losses into fast recovery (RFC8985). Note that
453 TLP requires RACK to function properly (see tcp_recovery below)
454
455 Possible values:
456
457 - 0 disables TLP
458 - 3 or 4 enables TLP
459
460 Default: 3
461
462tcp_ecn - INTEGER
463 Control use of Explicit Congestion Notification (ECN) by TCP.
464 ECN is used only when both ends of the TCP connection indicate support
465 for it. This feature is useful in avoiding losses due to congestion by
466 allowing supporting routers to signal congestion before having to drop
467 packets. A host that supports ECN both sends ECN at the IP layer and
468 feeds back ECN at the TCP layer. The highest variant of ECN feedback
469 that both peers support is chosen by the ECN negotiation (Accurate ECN,
470 ECN, or no ECN).
471
472 The highest negotiated variant for incoming connection requests
473 and the highest variant requested by outgoing connection
474 attempts:
475
476 ===== ==================== ====================
477 Value Incoming connections Outgoing connections
478 ===== ==================== ====================
479 0 No ECN No ECN
480 1 ECN ECN
481 2 ECN No ECN
482 3 AccECN AccECN
483 4 AccECN ECN
484 5 AccECN No ECN
485 ===== ==================== ====================
486
487 Default: 2
488
489tcp_ecn_option - INTEGER
490 Control Accurate ECN (AccECN) option sending when AccECN has been
491 successfully negotiated during handshake. Send logic inhibits
492 sending AccECN options regarless of this setting when no AccECN
493 option has been seen for the reverse direction.
494
495 Possible values are:
496
497 = ============================================================
498 0 Never send AccECN option. This also disables sending AccECN
499 option in SYN/ACK during handshake.
500 1 Send AccECN option sparingly according to the minimum option
501 rules outlined in draft-ietf-tcpm-accurate-ecn.
502 2 Send AccECN option on every packet whenever it fits into TCP
503 option space except when AccECN fallback is triggered.
504 3 Send AccECN option on every packet whenever it fits into TCP
505 option space even when AccECN fallback is triggered.
506 = ============================================================
507
508 Default: 2
509
510tcp_ecn_option_beacon - INTEGER
511 Control Accurate ECN (AccECN) option sending frequency per RTT and it
512 takes effect only when tcp_ecn_option is set to 2.
513
514 Default: 3 (AccECN will be send at least 3 times per RTT)
515
516tcp_ecn_fallback - BOOLEAN
517 If the kernel detects that ECN connection misbehaves, enable fall
518 back to non-ECN. Currently, this knob implements the fallback
519 from RFC3168, section 6.1.1.1., but we reserve that in future,
520 additional detection mechanisms could be implemented under this
521 knob. The value is not used, if tcp_ecn or per route (or congestion
522 control) ECN settings are disabled.
523
524 Possible values:
525
526 - 0 (disabled)
527 - 1 (enabled)
528
529 Default: 1 (enabled)
530
531tcp_fack - BOOLEAN
532 This is a legacy option, it has no effect anymore.
533
534tcp_fin_timeout - INTEGER
535 The length of time an orphaned (no longer referenced by any
536 application) connection will remain in the FIN_WAIT_2 state
537 before it is aborted at the local end. While a perfectly
538 valid "receive only" state for an un-orphaned connection, an
539 orphaned connection in FIN_WAIT_2 state could otherwise wait
540 forever for the remote to close its end of the connection.
541
542 Cf. tcp_max_orphans
543
544 Default: 60 seconds
545
546tcp_frto - INTEGER
547 Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
548 F-RTO is an enhanced recovery algorithm for TCP retransmission
549 timeouts. It is particularly beneficial in networks where the
550 RTT fluctuates (e.g., wireless). F-RTO is sender-side only
551 modification. It does not require any support from the peer.
552
553 By default it's enabled with a non-zero value. 0 disables F-RTO.
554
555tcp_fwmark_accept - BOOLEAN
556 If enabled, incoming connections to listening sockets that do not have a
557 socket mark will set the mark of the accepting socket to the fwmark of
558 the incoming SYN packet. This will cause all packets on that connection
559 (starting from the first SYNACK) to be sent with that fwmark. The
560 listening socket's mark is unchanged. Listening sockets that already
561 have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
562 unaffected.
563
564 Possible values:
565
566 - 0 (disabled)
567 - 1 (enabled)
568
569 Default: 0 (disabled)
570
571tcp_invalid_ratelimit - INTEGER
572 Limit the maximal rate for sending duplicate acknowledgments
573 in response to incoming TCP packets that are for an existing
574 connection but that are invalid due to any of these reasons:
575
576 (a) out-of-window sequence number,
577 (b) out-of-window acknowledgment number, or
578 (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
579
580 This can help mitigate simple "ack loop" DoS attacks, wherein
581 a buggy or malicious middlebox or man-in-the-middle can
582 rewrite TCP header fields in manner that causes each endpoint
583 to think that the other is sending invalid TCP segments, thus
584 causing each side to send an unterminating stream of duplicate
585 acknowledgments for invalid segments.
586
587 Using 0 disables rate-limiting of dupacks in response to
588 invalid segments; otherwise this value specifies the minimal
589 space between sending such dupacks, in milliseconds.
590
591 Default: 500 (milliseconds).
592
593tcp_keepalive_time - INTEGER
594 How often TCP sends out keepalive messages when keepalive is enabled.
595 Default: 2hours.
596
597tcp_keepalive_probes - INTEGER
598 How many keepalive probes TCP sends out, until it decides that the
599 connection is broken. Default value: 9.
600
601tcp_keepalive_intvl - INTEGER
602 How frequently the probes are send out. Multiplied by
603 tcp_keepalive_probes it is time to kill not responding connection,
604 after probes started. Default value: 75sec i.e. connection
605 will be aborted after ~11 minutes of retries.
606
607tcp_l3mdev_accept - BOOLEAN
608 Enables child sockets to inherit the L3 master device index.
609 Enabling this option allows a "global" listen socket to work
610 across L3 master domains (e.g., VRFs) with connected sockets
611 derived from the listen socket to be bound to the L3 domain in
612 which the packets originated. Only valid when the kernel was
613 compiled with CONFIG_NET_L3_MASTER_DEV.
614
615 Possible values:
616
617 - 0 (disabled)
618 - 1 (enabled)
619
620 Default: 0 (disabled)
621
622tcp_low_latency - BOOLEAN
623 This is a legacy option, it has no effect anymore.
624
625tcp_max_orphans - INTEGER
626 Maximal number of TCP sockets not attached to any user file handle,
627 held by system. If this number is exceeded orphaned connections are
628 reset immediately and warning is printed. This limit exists
629 only to prevent simple DoS attacks, you _must_ not rely on this
630 or lower the limit artificially, but rather increase it
631 (probably, after increasing installed memory),
632 if network conditions require more than default value,
633 and tune network services to linger and kill such states
634 more aggressively. Let me to remind again: each orphan eats
635 up to ~64K of unswappable memory.
636
637tcp_max_syn_backlog - INTEGER
638 Maximal number of remembered connection requests (SYN_RECV),
639 which have not received an acknowledgment from connecting client.
640
641 This is a per-listener limit.
642
643 The minimal value is 128 for low memory machines, and it will
644 increase in proportion to the memory of machine.
645
646 If server suffers from overload, try increasing this number.
647
648 Remember to also check /proc/sys/net/core/somaxconn
649 A SYN_RECV request socket consumes about 304 bytes of memory.
650
651tcp_max_tw_buckets - INTEGER
652 Maximal number of timewait sockets held by system simultaneously.
653 If this number is exceeded time-wait socket is immediately destroyed
654 and warning is printed. This limit exists only to prevent
655 simple DoS attacks, you _must_ not lower the limit artificially,
656 but rather increase it (probably, after increasing installed memory),
657 if network conditions require more than default value.
658
659tcp_mem - vector of 3 INTEGERs: min, pressure, max
660 min: below this number of pages TCP is not bothered about its
661 memory appetite.
662
663 pressure: when amount of memory allocated by TCP exceeds this number
664 of pages, TCP moderates its memory consumption and enters memory
665 pressure mode, which is exited when memory consumption falls
666 under "min".
667
668 max: number of pages allowed for queueing by all TCP sockets.
669
670 Defaults are calculated at boot time from amount of available
671 memory.
672
673tcp_min_rtt_wlen - INTEGER
674 The window length of the windowed min filter to track the minimum RTT.
675 A shorter window lets a flow more quickly pick up new (higher)
676 minimum RTT when it is moved to a longer path (e.g., due to traffic
677 engineering). A longer window makes the filter more resistant to RTT
678 inflations such as transient congestion. The unit is seconds.
679
680 Possible values: 0 - 86400 (1 day)
681
682 Default: 300
683
684tcp_moderate_rcvbuf - BOOLEAN
685 If enabled, TCP performs receive buffer auto-tuning, attempting to
686 automatically size the buffer (no greater than tcp_rmem[2]) to
687 match the size required by the path for full throughput.
688
689 Possible values:
690
691 - 0 (disabled)
692 - 1 (enabled)
693
694 Default: 1 (enabled)
695
696tcp_rcvbuf_low_rtt - INTEGER
697 rcvbuf autotuning can over estimate final socket rcvbuf, which
698 can lead to cache trashing for high throughput flows.
699
700 For small RTT flows (below tcp_rcvbuf_low_rtt usecs), we can relax
701 rcvbuf growth: Few additional ms to reach the final (and smaller)
702 rcvbuf is a good tradeoff.
703
704 Default : 1000 (1 ms)
705
706tcp_mtu_probing - INTEGER
707 Controls TCP Packetization-Layer Path MTU Discovery. Takes three
708 values:
709
710 - 0 - Disabled
711 - 1 - Disabled by default, enabled when an ICMP black hole detected
712 - 2 - Always enabled, use initial MSS of tcp_base_mss.
713
714tcp_probe_interval - UNSIGNED INTEGER
715 Controls how often to start TCP Packetization-Layer Path MTU
716 Discovery reprobe. The default is reprobing every 10 minutes as
717 per RFC4821.
718
719tcp_probe_threshold - INTEGER
720 Controls when TCP Packetization-Layer Path MTU Discovery probing
721 will stop in respect to the width of search range in bytes. Default
722 is 8 bytes.
723
724tcp_no_metrics_save - BOOLEAN
725 By default, TCP saves various connection metrics in the route cache
726 when the connection closes, so that connections established in the
727 near future can use these to set initial conditions. Usually, this
728 increases overall performance, but may sometimes cause performance
729 degradation. If enabled, TCP will not cache metrics on closing
730 connections.
731
732 Possible values:
733
734 - 0 (disabled)
735 - 1 (enabled)
736
737 Default: 0 (disabled)
738
739tcp_no_ssthresh_metrics_save - BOOLEAN
740 Controls whether TCP saves ssthresh metrics in the route cache.
741 If enabled, ssthresh metrics are disabled.
742
743 Possible values:
744
745 - 0 (disabled)
746 - 1 (enabled)
747
748 Default: 1 (enabled)
749
750tcp_orphan_retries - INTEGER
751 This value influences the timeout of a locally closed TCP connection,
752 when RTO retransmissions remain unacknowledged.
753 See tcp_retries2 for more details.
754
755 The default value is 8.
756
757 If your machine is a loaded WEB server,
758 you should think about lowering this value, such sockets
759 may consume significant resources. Cf. tcp_max_orphans.
760
761tcp_recovery - INTEGER
762 This value is a bitmap to enable various experimental loss recovery
763 features.
764
765 ========= =============================================================
766 RACK: 0x1 enables RACK loss detection, for fast detection of lost
767 retransmissions and tail drops, and resilience to
768 reordering. currently, setting this bit to 0 has no
769 effect, since RACK is the only supported loss detection
770 algorithm.
771
772 RACK: 0x2 makes RACK's reordering window static (min_rtt/4).
773
774 RACK: 0x4 disables RACK's DUPACK threshold heuristic
775 ========= =============================================================
776
777 Default: 0x1
778
779tcp_reflect_tos - BOOLEAN
780 For listening sockets, reuse the DSCP value of the initial SYN message
781 for outgoing packets. This allows to have both directions of a TCP
782 stream to use the same DSCP value, assuming DSCP remains unchanged for
783 the lifetime of the connection.
784
785 This options affects both IPv4 and IPv6.
786
787 Possible values:
788
789 - 0 (disabled)
790 - 1 (enabled)
791
792 Default: 0 (disabled)
793
794tcp_reordering - INTEGER
795 Initial reordering level of packets in a TCP stream.
796 TCP stack can then dynamically adjust flow reordering level
797 between this initial value and tcp_max_reordering
798
799 Default: 3
800
801tcp_max_reordering - INTEGER
802 Maximal reordering level of packets in a TCP stream.
803 300 is a fairly conservative value, but you might increase it
804 if paths are using per packet load balancing (like bonding rr mode)
805
806 Default: 300
807
808tcp_retrans_collapse - BOOLEAN
809 Bug-to-bug compatibility with some broken printers.
810 On retransmit try to send bigger packets to work around bugs in
811 certain TCP stacks.
812
813 Possible values:
814
815 - 0 (disabled)
816 - 1 (enabled)
817
818 Default: 1 (enabled)
819
820tcp_retries1 - INTEGER
821 This value influences the time, after which TCP decides, that
822 something is wrong due to unacknowledged RTO retransmissions,
823 and reports this suspicion to the network layer.
824 See tcp_retries2 for more details.
825
826 RFC 1122 recommends at least 3 retransmissions, which is the
827 default.
828
829tcp_retries2 - INTEGER
830 This value influences the timeout of an alive TCP connection,
831 when RTO retransmissions remain unacknowledged.
832 Given a value of N, a hypothetical TCP connection following
833 exponential backoff with an initial RTO of TCP_RTO_MIN would
834 retransmit N times before killing the connection at the (N+1)th RTO.
835
836 The default value of 15 yields a hypothetical timeout of 924.6
837 seconds and is a lower bound for the effective timeout.
838 TCP will effectively time out at the first RTO which exceeds the
839 hypothetical timeout.
840 If tcp_rto_max_ms is decreased, it is recommended to also
841 change tcp_retries2.
842
843 RFC 1122 recommends at least 100 seconds for the timeout,
844 which corresponds to a value of at least 8.
845
846tcp_rfc1337 - BOOLEAN
847 If enabled, the TCP stack behaves conforming to RFC1337. If unset,
848 we are not conforming to RFC, but prevent TCP TIME_WAIT
849 assassination.
850
851 Possible values:
852
853 - 0 (disabled)
854 - 1 (enabled)
855
856 Default: 0 (disabled)
857
858tcp_rmem - vector of 3 INTEGERs: min, default, max
859 min: Minimal size of receive buffer used by TCP sockets.
860 It is guaranteed to each TCP socket, even under moderate memory
861 pressure.
862
863 Default: 4K
864
865 default: initial size of receive buffer used by TCP sockets.
866 This value overrides net.core.rmem_default used by other protocols.
867 Default: 131072 bytes.
868 This value results in initial window of 65535.
869
870 max: maximal size of receive buffer allowed for automatically
871 selected receiver buffers for TCP socket.
872 Calling setsockopt() with SO_RCVBUF disables
873 automatic tuning of that socket's receive buffer size, in which
874 case this value is ignored.
875 Default: between 131072 and 32MB, depending on RAM size.
876
877tcp_sack - BOOLEAN
878 Enable select acknowledgments (SACKS).
879
880 Possible values:
881
882 - 0 (disabled)
883 - 1 (enabled)
884
885 Default: 1 (enabled)
886
887tcp_comp_sack_rtt_percent - INTEGER
888 Percentage of SRTT used for the compressed SACK feature.
889 See tcp_comp_sack_nr, tcp_comp_sack_delay_ns, tcp_comp_sack_slack_ns.
890
891 Possible values : 1 - 1000
892
893 Default : 33 %
894
895tcp_comp_sack_delay_ns - LONG INTEGER
896 TCP tries to reduce number of SACK sent, using a timer based
897 on tcp_comp_sack_rtt_percent of SRTT, capped by this sysctl
898 in nano seconds.
899 The default is 1ms, based on TSO autosizing period.
900
901 Default : 1,000,000 ns (1 ms)
902
903tcp_comp_sack_slack_ns - LONG INTEGER
904 This sysctl control the slack used when arming the
905 timer used by SACK compression. This gives extra time
906 for small RTT flows, and reduces system overhead by allowing
907 opportunistic reduction of timer interrupts.
908 Too big values might reduce goodput.
909
910 Default : 10,000 ns (10 us)
911
912tcp_comp_sack_nr - INTEGER
913 Max number of SACK that can be compressed.
914 Using 0 disables SACK compression.
915
916 Default : 44
917
918tcp_backlog_ack_defer - BOOLEAN
919 If enabled, user thread processing socket backlog tries sending
920 one ACK for the whole queue. This helps to avoid potential
921 long latencies at end of a TCP socket syscall.
922
923 Possible values:
924
925 - 0 (disabled)
926 - 1 (enabled)
927
928 Default: 1 (enabled)
929
930tcp_slow_start_after_idle - BOOLEAN
931 If enabled, provide RFC2861 behavior and time out the congestion
932 window after an idle period. An idle period is defined at
933 the current RTO. If unset, the congestion window will not
934 be timed out after an idle period.
935
936 Possible values:
937
938 - 0 (disabled)
939 - 1 (enabled)
940
941 Default: 1 (enabled)
942
943tcp_stdurg - BOOLEAN
944 Use the Host requirements interpretation of the TCP urgent pointer field.
945 Most hosts use the older BSD interpretation, so if enabled,
946 Linux might not communicate correctly with them.
947
948 Possible values:
949
950 - 0 (disabled)
951 - 1 (enabled)
952
953 Default: 0 (disabled)
954
955tcp_synack_retries - INTEGER
956 Number of times SYNACKs for a passive TCP connection attempt will
957 be retransmitted. Should not be higher than 255. Default value
958 is 5, which corresponds to 31seconds till the last retransmission
959 with the current initial RTO of 1second. With this the final timeout
960 for a passive TCP connection will happen after 63seconds.
961
962tcp_syncookies - INTEGER
963 Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
964 Send out syncookies when the syn backlog queue of a socket
965 overflows. This is to prevent against the common 'SYN flood attack'
966 Default: 1
967
968 Note, that syncookies is fallback facility.
969 It MUST NOT be used to help highly loaded servers to stand
970 against legal connection rate. If you see SYN flood warnings
971 in your logs, but investigation shows that they occur
972 because of overload with legal connections, you should tune
973 another parameters until this warning disappear.
974 See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
975
976 syncookies seriously violate TCP protocol, do not allow
977 to use TCP extensions, can result in serious degradation
978 of some services (f.e. SMTP relaying), visible not by you,
979 but your clients and relays, contacting you. While you see
980 SYN flood warnings in logs not being really flooded, your server
981 is seriously misconfigured.
982
983 If you want to test which effects syncookies have to your
984 network connections you can set this knob to 2 to enable
985 unconditionally generation of syncookies.
986
987tcp_migrate_req - BOOLEAN
988 The incoming connection is tied to a specific listening socket when
989 the initial SYN packet is received during the three-way handshake.
990 When a listener is closed, in-flight request sockets during the
991 handshake and established sockets in the accept queue are aborted.
992
993 If the listener has SO_REUSEPORT enabled, other listeners on the
994 same port should have been able to accept such connections. This
995 option makes it possible to migrate such child sockets to another
996 listener after close() or shutdown().
997
998 The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
999 usually be used to define the policy to pick an alive listener.
1000 Otherwise, the kernel will randomly pick an alive listener only if
1001 this option is enabled.
1002
1003 Note that migration between listeners with different settings may
1004 crash applications. Let's say migration happens from listener A to
1005 B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
1006 the requests migrated from A. To avoid such a situation, cancel
1007 migration by returning SK_DROP in the type of eBPF program, or
1008 disable this option.
1009
1010 Possible values:
1011
1012 - 0 (disabled)
1013 - 1 (enabled)
1014
1015 Default: 0 (disabled)
1016
1017tcp_fastopen - INTEGER
1018 Enable TCP Fast Open (RFC7413) to send and accept data in the opening
1019 SYN packet.
1020
1021 The client support is enabled by flag 0x1 (on by default). The client
1022 then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
1023 rather than connect() to send data in SYN.
1024
1025 The server support is enabled by flag 0x2 (off by default). Then
1026 either enable for all listeners with another flag (0x400) or
1027 enable individual listeners via TCP_FASTOPEN socket option with
1028 the option value being the length of the syn-data backlog.
1029
1030 The values (bitmap) are
1031
1032 ===== ======== ======================================================
1033 0x1 (client) enables sending data in the opening SYN on the client.
1034 0x2 (server) enables the server support, i.e., allowing data in
1035 a SYN packet to be accepted and passed to the
1036 application before 3-way handshake finishes.
1037 0x4 (client) send data in the opening SYN regardless of cookie
1038 availability and without a cookie option.
1039 0x200 (server) accept data-in-SYN w/o any cookie option present.
1040 0x400 (server) enable all listeners to support Fast Open by
1041 default without explicit TCP_FASTOPEN socket option.
1042 ===== ======== ======================================================
1043
1044 Default: 0x1
1045
1046 Note that additional client or server features are only
1047 effective if the basic support (0x1 and 0x2) are enabled respectively.
1048
1049tcp_fastopen_blackhole_timeout_sec - INTEGER
1050 Initial time period in second to disable Fastopen on active TCP sockets
1051 when a TFO firewall blackhole issue happens.
1052 This time period will grow exponentially when more blackhole issues
1053 get detected right after Fastopen is re-enabled and will reset to
1054 initial value when the blackhole issue goes away.
1055 0 to disable the blackhole detection.
1056
1057 By default, it is set to 0 (feature is disabled).
1058
1059tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
1060 The list consists of a primary key and an optional backup key. The
1061 primary key is used for both creating and validating cookies, while the
1062 optional backup key is only used for validating cookies. The purpose of
1063 the backup key is to maximize TFO validation when keys are rotated.
1064
1065 A randomly chosen primary key may be configured by the kernel if
1066 the tcp_fastopen sysctl is set to 0x400 (see above), or if the
1067 TCP_FASTOPEN setsockopt() optname is set and a key has not been
1068 previously configured via sysctl. If keys are configured via
1069 setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
1070 per-socket keys will be used instead of any keys that are specified via
1071 sysctl.
1072
1073 A key is specified as 4 8-digit hexadecimal integers which are separated
1074 by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
1075 omitted. A primary and a backup key may be specified by separating them
1076 by a comma. If only one key is specified, it becomes the primary key and
1077 any previously configured backup keys are removed.
1078
1079tcp_syn_retries - INTEGER
1080 Number of times initial SYNs for an active TCP connection attempt
1081 will be retransmitted. Should not be higher than 127. Default value
1082 is 6, which corresponds to 67seconds (with tcp_syn_linear_timeouts = 4)
1083 till the last retransmission with the current initial RTO of 1second.
1084 With this the final timeout for an active TCP connection attempt
1085 will happen after 131seconds.
1086
1087tcp_timestamps - INTEGER
1088 Enable timestamps as defined in RFC1323.
1089
1090 - 0: Disabled.
1091 - 1: Enable timestamps as defined in RFC1323 and use random offset for
1092 each connection rather than only using the current time.
1093 - 2: Like 1, but without random offsets.
1094
1095 Default: 1
1096
1097tcp_min_tso_segs - INTEGER
1098 Minimal number of segments per TSO frame.
1099
1100 Since linux-3.12, TCP does an automatic sizing of TSO frames,
1101 depending on flow rate, instead of filling 64Kbytes packets.
1102 For specific usages, it's possible to force TCP to build big
1103 TSO frames. Note that TCP stack might split too big TSO packets
1104 if available window is too small.
1105
1106 Default: 2
1107
1108tcp_tso_rtt_log - INTEGER
1109 Adjustment of TSO packet sizes based on min_rtt
1110
1111 Starting from linux-5.18, TCP autosizing can be tweaked
1112 for flows having small RTT.
1113
1114 Old autosizing was splitting the pacing budget to send 1024 TSO
1115 per second.
1116
1117 tso_packet_size = sk->sk_pacing_rate / 1024;
1118
1119 With the new mechanism, we increase this TSO sizing using:
1120
1121 distance = min_rtt_usec / (2^tcp_tso_rtt_log)
1122 tso_packet_size += gso_max_size >> distance;
1123
1124 This means that flows between very close hosts can use bigger
1125 TSO packets, reducing their cpu costs.
1126
1127 If you want to use the old autosizing, set this sysctl to 0.
1128
1129 Default: 9 (2^9 = 512 usec)
1130
1131tcp_pacing_ss_ratio - INTEGER
1132 sk->sk_pacing_rate is set by TCP stack using a ratio applied
1133 to current rate. (current_rate = cwnd * mss / srtt)
1134 If TCP is in slow start, tcp_pacing_ss_ratio is applied
1135 to let TCP probe for bigger speeds, assuming cwnd can be
1136 doubled every other RTT.
1137
1138 Default: 200
1139
1140tcp_pacing_ca_ratio - INTEGER
1141 sk->sk_pacing_rate is set by TCP stack using a ratio applied
1142 to current rate. (current_rate = cwnd * mss / srtt)
1143 If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
1144 is applied to conservatively probe for bigger throughput.
1145
1146 Default: 120
1147
1148tcp_syn_linear_timeouts - INTEGER
1149 The number of times for an active TCP connection to retransmit SYNs with
1150 a linear backoff timeout before defaulting to an exponential backoff
1151 timeout. This has no effect on SYNACK at the passive TCP side.
1152
1153 With an initial RTO of 1 and tcp_syn_linear_timeouts = 4 we would
1154 expect SYN RTOs to be: 1, 1, 1, 1, 1, 2, 4, ... (4 linear timeouts,
1155 and the first exponential backoff using 2^0 * initial_RTO).
1156 Default: 4
1157
1158tcp_tso_win_divisor - INTEGER
1159 This allows control over what percentage of the congestion window
1160 can be consumed by a single TSO frame.
1161 The setting of this parameter is a choice between burstiness and
1162 building larger TSO frames.
1163
1164 Default: 3
1165
1166tcp_tw_reuse - INTEGER
1167 Enable reuse of TIME-WAIT sockets for new connections when it is
1168 safe from protocol viewpoint.
1169
1170 - 0 - disable
1171 - 1 - global enable
1172 - 2 - enable for loopback traffic only
1173
1174 It should not be changed without advice/request of technical
1175 experts.
1176
1177 Default: 2
1178
1179tcp_tw_reuse_delay - UNSIGNED INTEGER
1180 The delay in milliseconds before a TIME-WAIT socket can be reused by a
1181 new connection, if TIME-WAIT socket reuse is enabled. The actual reuse
1182 threshold is within [N, N+1] range, where N is the requested delay in
1183 milliseconds, to ensure the delay interval is never shorter than the
1184 configured value.
1185
1186 This setting contains an assumption about the other TCP timestamp clock
1187 tick interval. It should not be set to a value lower than the peer's
1188 clock tick for PAWS (Protection Against Wrapped Sequence numbers)
1189 mechanism work correctly for the reused connection.
1190
1191 Default: 1000 (milliseconds)
1192
1193tcp_window_scaling - BOOLEAN
1194 Enable window scaling as defined in RFC1323.
1195
1196 Possible values:
1197
1198 - 0 (disabled)
1199 - 1 (enabled)
1200
1201 Default: 1 (enabled)
1202
1203tcp_shrink_window - BOOLEAN
1204 This changes how the TCP receive window is calculated.
1205
1206 RFC 7323, section 2.4, says there are instances when a retracted
1207 window can be offered, and that TCP implementations MUST ensure
1208 that they handle a shrinking window, as specified in RFC 1122.
1209
1210 Possible values:
1211
1212 - 0 (disabled) - The window is never shrunk.
1213 - 1 (enabled) - The window is shrunk when necessary to remain within
1214 the memory limit set by autotuning (sk_rcvbuf).
1215 This only occurs if a non-zero receive window
1216 scaling factor is also in effect.
1217
1218 Default: 0 (disabled)
1219
1220tcp_wmem - vector of 3 INTEGERs: min, default, max
1221 min: Amount of memory reserved for send buffers for TCP sockets.
1222 Each TCP socket has rights to use it due to fact of its birth.
1223
1224 Default: 4K
1225
1226 default: initial size of send buffer used by TCP sockets. This
1227 value overrides net.core.wmem_default used by other protocols.
1228
1229 It is usually lower than net.core.wmem_default.
1230
1231 Default: 16K
1232
1233 max: Maximal amount of memory allowed for automatically tuned
1234 send buffers for TCP sockets. This value does not override
1235 net.core.wmem_max. Calling setsockopt() with SO_SNDBUF disables
1236 automatic tuning of that socket's send buffer size, in which case
1237 this value is ignored.
1238
1239 Default: between 64K and 4MB, depending on RAM size.
1240
1241tcp_notsent_lowat - UNSIGNED INTEGER
1242 A TCP socket can control the amount of unsent bytes in its write queue,
1243 thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
1244 reports POLLOUT events if the amount of unsent bytes is below a per
1245 socket value, and if the write queue is not full. sendmsg() will
1246 also not add new buffers if the limit is hit.
1247
1248 This global variable controls the amount of unsent data for
1249 sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
1250 to the global variable has immediate effect.
1251
1252 Default: UINT_MAX (0xFFFFFFFF)
1253
1254tcp_workaround_signed_windows - BOOLEAN
1255 If enabled, assume no receipt of a window scaling option means the
1256 remote TCP is broken and treats the window as a signed quantity.
1257 If disabled, assume the remote TCP is not broken even if we do
1258 not receive a window scaling option from them.
1259
1260 Possible values:
1261
1262 - 0 (disabled)
1263 - 1 (enabled)
1264
1265 Default: 0 (disabled)
1266
1267tcp_thin_linear_timeouts - BOOLEAN
1268 Enable dynamic triggering of linear timeouts for thin streams.
1269 If enabled, a check is performed upon retransmission by timeout to
1270 determine if the stream is thin (less than 4 packets in flight).
1271 As long as the stream is found to be thin, up to 6 linear
1272 timeouts may be performed before exponential backoff mode is
1273 initiated. This improves retransmission latency for
1274 non-aggressive thin streams, often found to be time-dependent.
1275 For more information on thin streams, see
1276 Documentation/networking/tcp-thin.rst
1277
1278 Possible values:
1279
1280 - 0 (disabled)
1281 - 1 (enabled)
1282
1283 Default: 0 (disabled)
1284
1285tcp_limit_output_bytes - INTEGER
1286 Controls TCP Small Queue limit per tcp socket.
1287 TCP bulk sender tends to increase packets in flight until it
1288 gets losses notifications. With SNDBUF autotuning, this can
1289 result in a large amount of packets queued on the local machine
1290 (e.g.: qdiscs, CPU backlog, or device) hurting latency of other
1291 flows, for typical pfifo_fast qdiscs. tcp_limit_output_bytes
1292 limits the number of bytes on qdisc or device to reduce artificial
1293 RTT/cwnd and reduce bufferbloat.
1294
1295 Default: 4194304 (4 MB)
1296
1297tcp_challenge_ack_limit - INTEGER
1298 Limits number of Challenge ACK sent per second, as recommended
1299 in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
1300 Note that this per netns rate limit can allow some side channel
1301 attacks and probably should not be enabled.
1302 TCP stack implements per TCP socket limits anyway.
1303 Default: INT_MAX (unlimited)
1304
1305tcp_ehash_entries - INTEGER
1306 Show the number of hash buckets for TCP sockets in the current
1307 networking namespace.
1308
1309 A negative value means the networking namespace does not own its
1310 hash buckets and shares the initial networking namespace's one.
1311
1312tcp_child_ehash_entries - INTEGER
1313 Control the number of hash buckets for TCP sockets in the child
1314 networking namespace, which must be set before clone() or unshare().
1315
1316 If the value is not 0, the kernel uses a value rounded up to 2^n
1317 as the actual hash bucket size. 0 is a special value, meaning
1318 the child networking namespace will share the initial networking
1319 namespace's hash buckets.
1320
1321 Note that the child will use the global one in case the kernel
1322 fails to allocate enough memory. In addition, the global hash
1323 buckets are spread over available NUMA nodes, but the allocation
1324 of the child hash table depends on the current process's NUMA
1325 policy, which could result in performance differences.
1326
1327 Note also that the default value of tcp_max_tw_buckets and
1328 tcp_max_syn_backlog depend on the hash bucket size.
1329
1330 Possible values: 0, 2^n (n: 0 - 24 (16Mi))
1331
1332 Default: 0
1333
1334tcp_plb_enabled - BOOLEAN
1335 If enabled and the underlying congestion control (e.g. DCTCP) supports
1336 and enables PLB feature, TCP PLB (Protective Load Balancing) is
1337 enabled. PLB is described in the following paper:
1338 https://doi.org/10.1145/3544216.3544226. Based on PLB parameters,
1339 upon sensing sustained congestion, TCP triggers a change in
1340 flow label field for outgoing IPv6 packets. A change in flow label
1341 field potentially changes the path of outgoing packets for switches
1342 that use ECMP/WCMP for routing.
1343
1344 PLB changes socket txhash which results in a change in IPv6 Flow Label
1345 field, and currently no-op for IPv4 headers. It is possible
1346 to apply PLB for IPv4 with other network header fields (e.g. TCP
1347 or IPv4 options) or using encapsulation where outer header is used
1348 by switches to determine next hop. In either case, further host
1349 and switch side changes will be needed.
1350
1351 If enabled, PLB assumes that congestion signal (e.g. ECN) is made
1352 available and used by congestion control module to estimate a
1353 congestion measure (e.g. ce_ratio). PLB needs a congestion measure to
1354 make repathing decisions.
1355
1356 Possible values:
1357
1358 - 0 (disabled)
1359 - 1 (enabled)
1360
1361 Default: 0 (disabled)
1362
1363tcp_plb_idle_rehash_rounds - INTEGER
1364 Number of consecutive congested rounds (RTT) seen after which
1365 a rehash can be performed, given there are no packets in flight.
1366 This is referred to as M in PLB paper:
1367 https://doi.org/10.1145/3544216.3544226.
1368
1369 Possible Values: 0 - 31
1370
1371 Default: 3
1372
1373tcp_plb_rehash_rounds - INTEGER
1374 Number of consecutive congested rounds (RTT) seen after which
1375 a forced rehash can be performed. Be careful when setting this
1376 parameter, as a small value increases the risk of retransmissions.
1377 This is referred to as N in PLB paper:
1378 https://doi.org/10.1145/3544216.3544226.
1379
1380 Possible Values: 0 - 31
1381
1382 Default: 12
1383
1384tcp_plb_suspend_rto_sec - INTEGER
1385 Time, in seconds, to suspend PLB in event of an RTO. In order to avoid
1386 having PLB repath onto a connectivity "black hole", after an RTO a TCP
1387 connection suspends PLB repathing for a random duration between 1x and
1388 2x of this parameter. Randomness is added to avoid concurrent rehashing
1389 of multiple TCP connections. This should be set corresponding to the
1390 amount of time it takes to repair a failed link.
1391
1392 Possible Values: 0 - 255
1393
1394 Default: 60
1395
1396tcp_plb_cong_thresh - INTEGER
1397 Fraction of packets marked with congestion over a round (RTT) to
1398 tag that round as congested. This is referred to as K in the PLB paper:
1399 https://doi.org/10.1145/3544216.3544226.
1400
1401 The 0-1 fraction range is mapped to 0-256 range to avoid floating
1402 point operations. For example, 128 means that if at least 50% of
1403 the packets in a round were marked as congested then the round
1404 will be tagged as congested.
1405
1406 Setting threshold to 0 means that PLB repaths every RTT regardless
1407 of congestion. This is not intended behavior for PLB and should be
1408 used only for experimentation purpose.
1409
1410 Possible Values: 0 - 256
1411
1412 Default: 128
1413
1414tcp_pingpong_thresh - INTEGER
1415 The number of estimated data replies sent for estimated incoming data
1416 requests that must happen before TCP considers that a connection is a
1417 "ping-pong" (request-response) connection for which delayed
1418 acknowledgments can provide benefits.
1419
1420 This threshold is 1 by default, but some applications may need a higher
1421 threshold for optimal performance.
1422
1423 Possible Values: 1 - 255
1424
1425 Default: 1
1426
1427tcp_rto_min_us - INTEGER
1428 Minimal TCP retransmission timeout (in microseconds). Note that the
1429 rto_min route option has the highest precedence for configuring this
1430 setting, followed by the TCP_BPF_RTO_MIN and TCP_RTO_MIN_US socket
1431 options, followed by this tcp_rto_min_us sysctl.
1432
1433 The recommended practice is to use a value less or equal to 200000
1434 microseconds.
1435
1436 Possible Values: 1 - INT_MAX
1437
1438 Default: 200000
1439
1440tcp_rto_max_ms - INTEGER
1441 Maximal TCP retransmission timeout (in ms).
1442 Note that TCP_RTO_MAX_MS socket option has higher precedence.
1443
1444 When changing tcp_rto_max_ms, it is important to understand
1445 that tcp_retries2 might need a change.
1446
1447 Possible Values: 1000 - 120,000
1448
1449 Default: 120,000
1450
1451UDP variables
1452=============
1453
1454udp_l3mdev_accept - BOOLEAN
1455 Enabling this option allows a "global" bound socket to work
1456 across L3 master domains (e.g., VRFs) with packets capable of
1457 being received regardless of the L3 domain in which they
1458 originated. Only valid when the kernel was compiled with
1459 CONFIG_NET_L3_MASTER_DEV.
1460
1461 Possible values:
1462
1463 - 0 (disabled)
1464 - 1 (enabled)
1465
1466 Default: 0 (disabled)
1467
1468udp_mem - vector of 3 INTEGERs: min, pressure, max
1469 Number of pages allowed for queueing by all UDP sockets.
1470
1471 min: Number of pages allowed for queueing by all UDP sockets.
1472
1473 pressure: This value was introduced to follow format of tcp_mem.
1474
1475 max: This value was introduced to follow format of tcp_mem.
1476
1477 Default is calculated at boot time from amount of available memory.
1478
1479udp_rmem_min - INTEGER
1480 Minimal size of receive buffer used by UDP sockets in moderation.
1481 Each UDP socket is able to use the size for receiving data, even if
1482 total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1483
1484 Default: 4K
1485
1486udp_wmem_min - INTEGER
1487 UDP does not have tx memory accounting and this tunable has no effect.
1488
1489udp_hash_entries - INTEGER
1490 Show the number of hash buckets for UDP sockets in the current
1491 networking namespace.
1492
1493 A negative value means the networking namespace does not own its
1494 hash buckets and shares the initial networking namespace's one.
1495
1496udp_child_hash_entries - INTEGER
1497 Control the number of hash buckets for UDP sockets in the child
1498 networking namespace, which must be set before clone() or unshare().
1499
1500 If the value is not 0, the kernel uses a value rounded up to 2^n
1501 as the actual hash bucket size. 0 is a special value, meaning
1502 the child networking namespace will share the initial networking
1503 namespace's hash buckets.
1504
1505 Note that the child will use the global one in case the kernel
1506 fails to allocate enough memory. In addition, the global hash
1507 buckets are spread over available NUMA nodes, but the allocation
1508 of the child hash table depends on the current process's NUMA
1509 policy, which could result in performance differences.
1510
1511 Possible values: 0, 2^n (n: 7 (128) - 16 (64K))
1512
1513 Default: 0
1514
1515
1516RAW variables
1517=============
1518
1519raw_l3mdev_accept - BOOLEAN
1520 Enabling this option allows a "global" bound socket to work
1521 across L3 master domains (e.g., VRFs) with packets capable of
1522 being received regardless of the L3 domain in which they
1523 originated. Only valid when the kernel was compiled with
1524 CONFIG_NET_L3_MASTER_DEV.
1525
1526 Possible values:
1527
1528 - 0 (disabled)
1529 - 1 (enabled)
1530
1531 Default: 1 (enabled)
1532
1533CIPSOv4 Variables
1534=================
1535
1536cipso_cache_enable - BOOLEAN
1537 If enabled, enable additions to and lookups from the CIPSO label mapping
1538 cache. If disabled, additions are ignored and lookups always result in a
1539 miss. However, regardless of the setting the cache is still
1540 invalidated when required when means you can safely toggle this on and
1541 off and the cache will always be "safe".
1542
1543 Possible values:
1544
1545 - 0 (disabled)
1546 - 1 (enabled)
1547
1548 Default: 1 (enabled)
1549
1550cipso_cache_bucket_size - INTEGER
1551 The CIPSO label cache consists of a fixed size hash table with each
1552 hash bucket containing a number of cache entries. This variable limits
1553 the number of entries in each hash bucket; the larger the value is, the
1554 more CIPSO label mappings that can be cached. When the number of
1555 entries in a given hash bucket reaches this limit adding new entries
1556 causes the oldest entry in the bucket to be removed to make room.
1557
1558 Default: 10
1559
1560cipso_rbm_optfmt - BOOLEAN
1561 Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1562 the CIPSO draft specification (see Documentation/netlabel for details).
1563 This means that when set the CIPSO tag will be padded with empty
1564 categories in order to make the packet data 32-bit aligned.
1565
1566 Possible values:
1567
1568 - 0 (disabled)
1569 - 1 (enabled)
1570
1571 Default: 0 (disabled)
1572
1573cipso_rbm_strictvalid - BOOLEAN
1574 If enabled, do a very strict check of the CIPSO option when
1575 ip_options_compile() is called. If disabled, relax the checks done during
1576 ip_options_compile(). Either way is "safe" as errors are caught else
1577 where in the CIPSO processing code but setting this to 0 (False) should
1578 result in less work (i.e. it should be faster) but could cause problems
1579 with other implementations that require strict checking.
1580
1581 Possible values:
1582
1583 - 0 (disabled)
1584 - 1 (enabled)
1585
1586 Default: 0 (disabled)
1587
1588IP Variables
1589============
1590
1591ip_local_port_range - 2 INTEGERS
1592 Defines the local port range that is used by TCP and UDP to
1593 choose the local port. The first number is the first, the
1594 second the last local port number.
1595 If possible, it is better these numbers have different parity
1596 (one even and one odd value).
1597 Must be greater than or equal to ip_unprivileged_port_start.
1598 The default values are 32768 and 60999 respectively.
1599
1600ip_local_reserved_ports - list of comma separated ranges
1601 Specify the ports which are reserved for known third-party
1602 applications. These ports will not be used by automatic port
1603 assignments (e.g. when calling connect() or bind() with port
1604 number 0). Explicit port allocation behavior is unchanged.
1605
1606 The format used for both input and output is a comma separated
1607 list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1608 10). Writing to the file will clear all previously reserved
1609 ports and update the current list with the one given in the
1610 input.
1611
1612 Note that ip_local_port_range and ip_local_reserved_ports
1613 settings are independent and both are considered by the kernel
1614 when determining which ports are available for automatic port
1615 assignments.
1616
1617 You can reserve ports which are not in the current
1618 ip_local_port_range, e.g.::
1619
1620 $ cat /proc/sys/net/ipv4/ip_local_port_range
1621 32000 60999
1622 $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1623 8080,9148
1624
1625 although this is redundant. However such a setting is useful
1626 if later the port range is changed to a value that will
1627 include the reserved ports. Also keep in mind, that overlapping
1628 of these ranges may affect probability of selecting ephemeral
1629 ports which are right after block of reserved ports.
1630
1631 Default: Empty
1632
1633ip_local_port_step_width - INTEGER
1634 Defines the numerical maximum increment between successive port
1635 allocations within the ephemeral port range when an unavailable port is
1636 reached. This can be used to mitigate accumulated nodes in port
1637 distribution when reserved ports have been configured. Please note that
1638 port collisions may be more frequent in a system with a very high load.
1639
1640 It is recommended to set this value strictly larger than the largest
1641 contiguous block of ports configure in ip_local_reserved_ports. For
1642 large reserved port ranges, setting this to 3x or 4x the size of the
1643 largest block is advised. Using a value equal or greater than the local
1644 port range size completely solves the uneven port distribution problem,
1645 but it can degrade performance under port exhaustion situations.
1646
1647 Default: 0 (disabled)
1648
1649ip_unprivileged_port_start - INTEGER
1650 This is a per-namespace sysctl. It defines the first
1651 unprivileged port in the network namespace. Privileged ports
1652 require root or CAP_NET_BIND_SERVICE in order to bind to them.
1653 To disable all privileged ports, set this to 0. They must not
1654 overlap with the ip_local_port_range.
1655
1656 Default: 1024
1657
1658ip_nonlocal_bind - BOOLEAN
1659 If enabled, allows processes to bind() to non-local IP addresses,
1660 which can be quite useful - but may break some applications.
1661
1662 Possible values:
1663
1664 - 0 (disabled)
1665 - 1 (enabled)
1666
1667 Default: 0 (disabled)
1668
1669ip_autobind_reuse - BOOLEAN
1670 By default, bind() does not select the ports automatically even if
1671 the new socket and all sockets bound to the port have SO_REUSEADDR.
1672 ip_autobind_reuse allows bind() to reuse the port and this is useful
1673 when you use bind()+connect(), but may break some applications.
1674 The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1675 option should only be set by experts.
1676
1677 Possible values:
1678
1679 - 0 (disabled)
1680 - 1 (enabled)
1681
1682 Default: 0 (disabled)
1683
1684ip_dynaddr - INTEGER
1685 If set non-zero, enables support for dynamic addresses.
1686 If set to a non-zero value larger than 1, a kernel log
1687 message will be printed when dynamic address rewriting
1688 occurs.
1689
1690 Default: 0
1691
1692ip_early_demux - BOOLEAN
1693 Optimize input packet processing down to one demux for
1694 certain kinds of local sockets. Currently we only do this
1695 for established TCP and connected UDP sockets.
1696
1697 It may add an additional cost for pure routing workloads that
1698 reduces overall throughput, in such case you should disable it.
1699
1700 Possible values:
1701
1702 - 0 (disabled)
1703 - 1 (enabled)
1704
1705 Default: 1 (enabled)
1706
1707ping_group_range - 2 INTEGERS
1708 Restrict ICMP_PROTO datagram sockets to users in the group range.
1709 The default is "1 0", meaning, that nobody (not even root) may
1710 create ping sockets. Setting it to "100 100" would grant permissions
1711 to the single group. "0 4294967294" would enable it for the world, "100
1712 4294967294" would enable it for the users, but not daemons.
1713
1714tcp_early_demux - BOOLEAN
1715 Enable early demux for established TCP sockets.
1716
1717 Possible values:
1718
1719 - 0 (disabled)
1720 - 1 (enabled)
1721
1722 Default: 1 (enabled)
1723
1724udp_early_demux - BOOLEAN
1725 Enable early demux for connected UDP sockets. Disable this if
1726 your system could experience more unconnected load.
1727
1728 Possible values:
1729
1730 - 0 (disabled)
1731 - 1 (enabled)
1732
1733 Default: 1 (enabled)
1734
1735icmp_echo_ignore_all - BOOLEAN
1736 If enabled, then the kernel will ignore all ICMP ECHO
1737 requests sent to it.
1738
1739 Possible values:
1740
1741 - 0 (disabled)
1742 - 1 (enabled)
1743
1744 Default: 0 (disabled)
1745
1746icmp_echo_enable_probe - BOOLEAN
1747 If enabled, then the kernel will respond to RFC 8335 PROBE
1748 requests sent to it.
1749
1750 Possible values:
1751
1752 - 0 (disabled)
1753 - 1 (enabled)
1754
1755 Default: 0 (disabled)
1756
1757icmp_echo_ignore_broadcasts - BOOLEAN
1758 If enabled, then the kernel will ignore all ICMP ECHO and
1759 TIMESTAMP requests sent to it via broadcast/multicast.
1760
1761 Possible values:
1762
1763 - 0 (disabled)
1764 - 1 (enabled)
1765
1766 Default: 1 (enabled)
1767
1768icmp_ratelimit - INTEGER
1769 Limit the maximal rates for sending ICMP packets whose type matches
1770 icmp_ratemask (see below) to specific targets.
1771 0 to disable any limiting,
1772 otherwise the minimal space between responses in milliseconds.
1773 Note that another sysctl, icmp_msgs_per_sec limits the number
1774 of ICMP packets sent on all targets.
1775
1776 Default: 1000
1777
1778icmp_msgs_per_sec - INTEGER
1779 Limit maximal number of ICMP packets sent per second from this host.
1780 Only messages whose type matches icmp_ratemask (see below) are
1781 controlled by this limit. For security reasons, the precise count
1782 of messages per second is randomized.
1783
1784 Default: 10000
1785
1786icmp_msgs_burst - INTEGER
1787 icmp_msgs_per_sec controls number of ICMP packets sent per second,
1788 while icmp_msgs_burst controls the token bucket size.
1789 For security reasons, the precise burst size is randomized.
1790
1791 Default: 10000
1792
1793icmp_ratemask - INTEGER
1794 Mask made of ICMP types for which rates are being limited.
1795
1796 Significant bits: IHGFEDCBA9876543210
1797
1798 Default mask: 0000001100000011000 (6168)
1799
1800 Bit definitions (see include/linux/icmp.h):
1801
1802 = =========================
1803 0 Echo Reply
1804 3 Destination Unreachable [1]_
1805 4 Source Quench [1]_
1806 5 Redirect
1807 8 Echo Request
1808 B Time Exceeded [1]_
1809 C Parameter Problem [1]_
1810 D Timestamp Request
1811 E Timestamp Reply
1812 F Info Request
1813 G Info Reply
1814 H Address Mask Request
1815 I Address Mask Reply
1816 = =========================
1817
1818 .. [1] These are rate limited by default (see default mask above)
1819
1820icmp_ignore_bogus_error_responses - BOOLEAN
1821 Some routers violate RFC1122 by sending bogus responses to broadcast
1822 frames. Such violations are normally logged via a kernel warning.
1823 If enabled, the kernel will not give such warnings, which
1824 will avoid log file clutter.
1825
1826 Possible values:
1827
1828 - 0 (disabled)
1829 - 1 (enabled)
1830
1831 Default: 1 (enabled)
1832
1833icmp_errors_use_inbound_ifaddr - BOOLEAN
1834
1835 If disabled, icmp error messages are sent with the primary address of
1836 the exiting interface.
1837
1838 If enabled, the message will be sent with the primary address of
1839 the interface that received the packet that caused the icmp error.
1840 This is the behaviour many network administrators will expect from
1841 a router. And it can make debugging complicated network layouts
1842 much easier.
1843
1844 Note that if no primary address exists for the interface selected,
1845 then the primary address of the first non-loopback interface that
1846 has one will be used regardless of this setting.
1847
1848 Possible values:
1849
1850 - 0 (disabled)
1851 - 1 (enabled)
1852
1853 Default: 0 (disabled)
1854
1855icmp_errors_extension_mask - UNSIGNED INTEGER
1856 Bitmask of ICMP extensions to append to ICMPv4 error messages
1857 ("Destination Unreachable", "Time Exceeded" and "Parameter Problem").
1858 The original datagram is trimmed / padded to 128 bytes in order to be
1859 compatible with applications that do not comply with RFC 4884.
1860
1861 Possible extensions are:
1862
1863 ==== ==============================================================
1864 0x01 Incoming IP interface information according to RFC 5837.
1865 Extension will include the index, IPv4 address (if present),
1866 name and MTU of the IP interface that received the datagram
1867 which elicited the ICMP error.
1868 ==== ==============================================================
1869
1870 Default: 0x00 (no extensions)
1871
1872igmp_max_memberships - INTEGER
1873 Change the maximum number of multicast groups we can subscribe to.
1874 Default: 20
1875
1876 Theoretical maximum value is bounded by having to send a membership
1877 report in a single datagram (i.e. the report can't span multiple
1878 datagrams, or risk confusing the switch and leaving groups you don't
1879 intend to).
1880
1881 The number of supported groups 'M' is bounded by the number of group
1882 report entries you can fit into a single datagram of 65535 bytes.
1883
1884 M = 65536-sizeof (ip header)/(sizeof(Group record))
1885
1886 Group records are variable length, with a minimum of 12 bytes.
1887 So net.ipv4.igmp_max_memberships should not be set higher than:
1888
1889 (65536-24) / 12 = 5459
1890
1891 The value 5459 assumes no IP header options, so in practice
1892 this number may be lower.
1893
1894igmp_max_msf - INTEGER
1895 Maximum number of addresses allowed in the source filter list for a
1896 multicast group.
1897
1898 Default: 10
1899
1900igmp_qrv - INTEGER
1901 Controls the IGMP query robustness variable (see RFC2236 8.1).
1902
1903 Default: 2 (as specified by RFC2236 8.1)
1904
1905 Minimum: 1 (as specified by RFC6636 4.5)
1906
1907force_igmp_version - INTEGER
1908 - 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1909 allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1910 Present timer expires.
1911 - 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1912 receive IGMPv2/v3 query.
1913 - 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1914 IGMPv1 query message. Will reply report if receive IGMPv3 query.
1915 - 3 - Enforce to use IGMP version 3. The same react with default 0.
1916
1917 .. note::
1918
1919 this is not the same with force_mld_version because IGMPv3 RFC3376
1920 Security Considerations does not have clear description that we could
1921 ignore other version messages completely as MLDv2 RFC3810. So make
1922 this value as default 0 is recommended.
1923
1924``conf/interface/*``
1925 changes special settings per interface (where
1926 interface" is the name of your network interface)
1927
1928``conf/all/*``
1929 is special, changes the settings for all interfaces
1930
1931log_martians - BOOLEAN
1932 Log packets with impossible addresses to kernel log.
1933 log_martians for the interface will be enabled if at least one of
1934 conf/{all,interface}/log_martians is set to TRUE,
1935 it will be disabled otherwise
1936
1937accept_redirects - BOOLEAN
1938 Accept ICMP redirect messages.
1939 accept_redirects for the interface will be enabled if:
1940
1941 - both conf/{all,interface}/accept_redirects are TRUE in the case
1942 forwarding for the interface is enabled
1943
1944 or
1945
1946 - at least one of conf/{all,interface}/accept_redirects is TRUE in the
1947 case forwarding for the interface is disabled
1948
1949 accept_redirects for the interface will be disabled otherwise
1950
1951 default:
1952
1953 - TRUE (host)
1954 - FALSE (router)
1955
1956forwarding - BOOLEAN
1957 Enable IP forwarding on this interface. This controls whether packets
1958 received _on_ this interface can be forwarded.
1959
1960mc_forwarding - BOOLEAN
1961 Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1962 and a multicast routing daemon is required.
1963 conf/all/mc_forwarding must also be set to TRUE to enable multicast
1964 routing for the interface
1965
1966medium_id - INTEGER
1967 Integer value used to differentiate the devices by the medium they
1968 are attached to. Two devices can have different id values when
1969 the broadcast packets are received only on one of them.
1970 The default value 0 means that the device is the only interface
1971 to its medium, value of -1 means that medium is not known.
1972
1973 Currently, it is used to change the proxy_arp behavior:
1974 the proxy_arp feature is enabled for packets forwarded between
1975 two devices attached to different media.
1976
1977proxy_arp - BOOLEAN
1978 Do proxy arp.
1979
1980 proxy_arp for the interface will be enabled if at least one of
1981 conf/{all,interface}/proxy_arp is set to TRUE,
1982 it will be disabled otherwise
1983
1984proxy_arp_pvlan - BOOLEAN
1985 Private VLAN proxy arp.
1986
1987 Basically allow proxy arp replies back to the same interface
1988 (from which the ARP request/solicitation was received).
1989
1990 This is done to support (ethernet) switch features, like RFC
1991 3069, where the individual ports are NOT allowed to
1992 communicate with each other, but they are allowed to talk to
1993 the upstream router. As described in RFC 3069, it is possible
1994 to allow these hosts to communicate through the upstream
1995 router by proxy_arp'ing. Don't need to be used together with
1996 proxy_arp.
1997
1998 This technology is known by different names:
1999
2000 - In RFC 3069 it is called VLAN Aggregation.
2001 - Cisco and Allied Telesyn call it Private VLAN.
2002 - Hewlett-Packard call it Source-Port filtering or port-isolation.
2003 - Ericsson call it MAC-Forced Forwarding (RFC Draft).
2004
2005proxy_delay - INTEGER
2006 Delay proxy response.
2007
2008 Delay response to a neighbor solicitation when proxy_arp
2009 or proxy_ndp is enabled. A random value between [0, proxy_delay)
2010 will be chosen, setting to zero means reply with no delay.
2011 Value in jiffies. Defaults to 80.
2012
2013shared_media - BOOLEAN
2014 Send(router) or accept(host) RFC1620 shared media redirects.
2015 Overrides secure_redirects.
2016
2017 shared_media for the interface will be enabled if at least one of
2018 conf/{all,interface}/shared_media is set to TRUE,
2019 it will be disabled otherwise
2020
2021 default TRUE
2022
2023secure_redirects - BOOLEAN
2024 Accept ICMP redirect messages only to gateways listed in the
2025 interface's current gateway list. Even if disabled, RFC1122 redirect
2026 rules still apply.
2027
2028 Overridden by shared_media.
2029
2030 secure_redirects for the interface will be enabled if at least one of
2031 conf/{all,interface}/secure_redirects is set to TRUE,
2032 it will be disabled otherwise
2033
2034 default TRUE
2035
2036send_redirects - BOOLEAN
2037 Send redirects, if router.
2038
2039 send_redirects for the interface will be enabled if at least one of
2040 conf/{all,interface}/send_redirects is set to TRUE,
2041 it will be disabled otherwise
2042
2043 Default: TRUE
2044
2045bootp_relay - BOOLEAN
2046 Accept packets with source address 0.b.c.d destined
2047 not to this host as local ones. It is supposed, that
2048 BOOTP relay daemon will catch and forward such packets.
2049 conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
2050 for the interface
2051
2052 default FALSE
2053
2054 Not Implemented Yet.
2055
2056accept_source_route - BOOLEAN
2057 Accept packets with SRR option.
2058 conf/all/accept_source_route must also be set to TRUE to accept packets
2059 with SRR option on the interface
2060
2061 default
2062
2063 - TRUE (router)
2064 - FALSE (host)
2065
2066accept_local - BOOLEAN
2067 Accept packets with local source addresses. In combination with
2068 suitable routing, this can be used to direct packets between two
2069 local interfaces over the wire and have them accepted properly.
2070 default FALSE
2071
2072route_localnet - BOOLEAN
2073 Do not consider loopback addresses as martian source or destination
2074 while routing. This enables the use of 127/8 for local routing purposes.
2075
2076 default FALSE
2077
2078rp_filter - INTEGER
2079 - 0 - No source validation.
2080 - 1 - Strict mode as defined in RFC3704 Strict Reverse Path
2081 Each incoming packet is tested against the FIB and if the interface
2082 is not the best reverse path the packet check will fail.
2083 By default failed packets are discarded.
2084 - 2 - Loose mode as defined in RFC3704 Loose Reverse Path
2085 Each incoming packet's source address is also tested against the FIB
2086 and if the source address is not reachable via any interface
2087 the packet check will fail.
2088
2089 Current recommended practice in RFC3704 is to enable strict mode
2090 to prevent IP spoofing from DDos attacks. If using asymmetric routing
2091 or other complicated routing, then loose mode is recommended.
2092
2093 The max value from conf/{all,interface}/rp_filter is used
2094 when doing source validation on the {interface}.
2095
2096 Default value is 0. Note that some distributions enable it
2097 in startup scripts.
2098
2099src_valid_mark - BOOLEAN
2100 - 0 - The fwmark of the packet is not included in reverse path
2101 route lookup. This allows for asymmetric routing configurations
2102 utilizing the fwmark in only one direction, e.g., transparent
2103 proxying.
2104
2105 - 1 - The fwmark of the packet is included in reverse path route
2106 lookup. This permits rp_filter to function when the fwmark is
2107 used for routing traffic in both directions.
2108
2109 This setting also affects the utilization of fmwark when
2110 performing source address selection for ICMP replies, or
2111 determining addresses stored for the IPOPT_TS_TSANDADDR and
2112 IPOPT_RR IP options.
2113
2114 The max value from conf/{all,interface}/src_valid_mark is used.
2115
2116 Default value is 0.
2117
2118arp_filter - BOOLEAN
2119 - 1 - Allows you to have multiple network interfaces on the same
2120 subnet, and have the ARPs for each interface be answered
2121 based on whether or not the kernel would route a packet from
2122 the ARP'd IP out that interface (therefore you must use source
2123 based routing for this to work). In other words it allows control
2124 of which cards (usually 1) will respond to an arp request.
2125
2126 - 0 - (default) The kernel can respond to arp requests with addresses
2127 from other interfaces. This may seem wrong but it usually makes
2128 sense, because it increases the chance of successful communication.
2129 IP addresses are owned by the complete host on Linux, not by
2130 particular interfaces. Only for more complex setups like load-
2131 balancing, does this behaviour cause problems.
2132
2133 arp_filter for the interface will be enabled if at least one of
2134 conf/{all,interface}/arp_filter is set to TRUE,
2135 it will be disabled otherwise
2136
2137arp_announce - INTEGER
2138 Define different restriction levels for announcing the local
2139 source IP address from IP packets in ARP requests sent on
2140 interface:
2141
2142 - 0 - (default) Use any local address, configured on any interface
2143 - 1 - Try to avoid local addresses that are not in the target's
2144 subnet for this interface. This mode is useful when target
2145 hosts reachable via this interface require the source IP
2146 address in ARP requests to be part of their logical network
2147 configured on the receiving interface. When we generate the
2148 request we will check all our subnets that include the
2149 target IP and will preserve the source address if it is from
2150 such subnet. If there is no such subnet we select source
2151 address according to the rules for level 2.
2152 - 2 - Always use the best local address for this target.
2153 In this mode we ignore the source address in the IP packet
2154 and try to select local address that we prefer for talks with
2155 the target host. Such local address is selected by looking
2156 for primary IP addresses on all our subnets on the outgoing
2157 interface that include the target IP address. If no suitable
2158 local address is found we select the first local address
2159 we have on the outgoing interface or on all other interfaces,
2160 with the hope we will receive reply for our request and
2161 even sometimes no matter the source IP address we announce.
2162
2163 The max value from conf/{all,interface}/arp_announce is used.
2164
2165 Increasing the restriction level gives more chance for
2166 receiving answer from the resolved target while decreasing
2167 the level announces more valid sender's information.
2168
2169arp_ignore - INTEGER
2170 Define different modes for sending replies in response to
2171 received ARP requests that resolve local target IP addresses:
2172
2173 - 0 - (default): reply for any local target IP address, configured
2174 on any interface
2175 - 1 - reply only if the target IP address is local address
2176 configured on the incoming interface
2177 - 2 - reply only if the target IP address is local address
2178 configured on the incoming interface and both with the
2179 sender's IP address are part from same subnet on this interface
2180 - 3 - do not reply for local addresses configured with scope host,
2181 only resolutions for global and link addresses are replied
2182 - 4-7 - reserved
2183 - 8 - do not reply for all local addresses
2184
2185 The max value from conf/{all,interface}/arp_ignore is used
2186 when ARP request is received on the {interface}
2187
2188arp_notify - BOOLEAN
2189 Define mode for notification of address and device changes.
2190
2191 == ==========================================================
2192 0 (default): do nothing
2193 1 Generate gratuitous arp requests when device is brought up
2194 or hardware address changes.
2195 == ==========================================================
2196
2197arp_accept - INTEGER
2198 Define behavior for accepting gratuitous ARP (garp) frames from devices
2199 that are not already present in the ARP table:
2200
2201 - 0 - don't create new entries in the ARP table
2202 - 1 - create new entries in the ARP table
2203 - 2 - create new entries only if the source IP address is in the same
2204 subnet as an address configured on the interface that received the
2205 garp message.
2206
2207 Both replies and requests type gratuitous arp will trigger the
2208 ARP table to be updated, if this setting is on.
2209
2210 If the ARP table already contains the IP address of the
2211 gratuitous arp frame, the arp table will be updated regardless
2212 if this setting is on or off.
2213
2214arp_evict_nocarrier - BOOLEAN
2215 Clears the ARP cache on NOCARRIER events. This option is important for
2216 wireless devices where the ARP cache should not be cleared when roaming
2217 between access points on the same network. In most cases this should
2218 remain as the default (1).
2219
2220 Possible values:
2221
2222 - 0 (disabled) - Do not clear ARP cache on NOCARRIER events
2223 - 1 (enabled) - Clear the ARP cache on NOCARRIER events
2224
2225 Default: 1 (enabled)
2226
2227mcast_solicit - INTEGER
2228 The maximum number of multicast probes in INCOMPLETE state,
2229 when the associated hardware address is unknown. Defaults
2230 to 3.
2231
2232ucast_solicit - INTEGER
2233 The maximum number of unicast probes in PROBE state, when
2234 the hardware address is being reconfirmed. Defaults to 3.
2235
2236app_solicit - INTEGER
2237 The maximum number of probes to send to the user space ARP daemon
2238 via netlink before dropping back to multicast probes (see
2239 mcast_resolicit). Defaults to 0.
2240
2241mcast_resolicit - INTEGER
2242 The maximum number of multicast probes after unicast and
2243 app probes in PROBE state. Defaults to 0.
2244
2245disable_policy - BOOLEAN
2246 Disable IPSEC policy (SPD) for this interface
2247
2248 Possible values:
2249
2250 - 0 (disabled)
2251 - 1 (enabled)
2252
2253 Default: 0 (disabled)
2254
2255disable_xfrm - BOOLEAN
2256 Disable IPSEC encryption on this interface, whatever the policy
2257
2258 Possible values:
2259
2260 - 0 (disabled)
2261 - 1 (enabled)
2262
2263 Default: 0 (disabled)
2264
2265igmpv2_unsolicited_report_interval - INTEGER
2266 The interval in milliseconds in which the next unsolicited
2267 IGMPv1 or IGMPv2 report retransmit will take place.
2268
2269 Default: 10000 (10 seconds)
2270
2271igmpv3_unsolicited_report_interval - INTEGER
2272 The interval in milliseconds in which the next unsolicited
2273 IGMPv3 report retransmit will take place.
2274
2275 Default: 1000 (1 seconds)
2276
2277ignore_routes_with_linkdown - BOOLEAN
2278 Ignore routes whose link is down when performing a FIB lookup.
2279
2280 Possible values:
2281
2282 - 0 (disabled)
2283 - 1 (enabled)
2284
2285 Default: 0 (disabled)
2286
2287promote_secondaries - BOOLEAN
2288 When a primary IP address is removed from this interface
2289 promote a corresponding secondary IP address instead of
2290 removing all the corresponding secondary IP addresses.
2291
2292 Possible values:
2293
2294 - 0 (disabled)
2295 - 1 (enabled)
2296
2297 Default: 0 (disabled)
2298
2299drop_unicast_in_l2_multicast - BOOLEAN
2300 Drop any unicast IP packets that are received in link-layer
2301 multicast (or broadcast) frames.
2302
2303 This behavior (for multicast) is actually a SHOULD in RFC
2304 1122, but is disabled by default for compatibility reasons.
2305
2306 Possible values:
2307
2308 - 0 (disabled)
2309 - 1 (enabled)
2310
2311 Default: 0 (disabled)
2312
2313drop_gratuitous_arp - BOOLEAN
2314 Drop all gratuitous ARP frames, for example if there's a known
2315 good ARP proxy on the network and such frames need not be used
2316 (or in the case of 802.11, must not be used to prevent attacks.)
2317
2318 Possible values:
2319
2320 - 0 (disabled)
2321 - 1 (enabled)
2322
2323 Default: 0 (disabled)
2324
2325
2326tag - INTEGER
2327 Allows you to write a number, which can be used as required.
2328
2329 Default value is 0.
2330
2331xfrm4_gc_thresh - INTEGER
2332 (Obsolete since linux-4.14)
2333 The threshold at which we will start garbage collecting for IPv4
2334 destination cache entries. At twice this value the system will
2335 refuse new allocations.
2336
2337igmp_link_local_mcast_reports - BOOLEAN
2338 Enable IGMP reports for link local multicast groups in the
2339 224.0.0.X range.
2340
2341 Default TRUE
2342
2343Alexey Kuznetsov.
2344kuznet@ms2.inr.ac.ru
2345
2346Updated by:
2347
2348- Andi Kleen
2349 ak@muc.de
2350- Nicolas Delon
2351 delon.nicolas@wanadoo.fr
2352
2353
2354
2355
2356/proc/sys/net/ipv6/* Variables
2357==============================
2358
2359IPv6 has no global variables such as tcp_*. tcp_* settings under ipv4/ also
2360apply to IPv6 [XXX?].
2361
2362bindv6only - BOOLEAN
2363 Default value for IPV6_V6ONLY socket option,
2364 which restricts use of the IPv6 socket to IPv6 communication
2365 only.
2366
2367 Possible values:
2368
2369 - 0 (disabled) - enable IPv4-mapped address feature
2370 - 1 (enabled) - disable IPv4-mapped address feature
2371
2372 Default: 0 (disabled)
2373
2374flowlabel_consistency - BOOLEAN
2375 Protect the consistency (and unicity) of flow label.
2376 You have to disable it to use IPV6_FL_F_REFLECT flag on the
2377 flow label manager.
2378
2379 Possible values:
2380
2381 - 0 (disabled)
2382 - 1 (enabled)
2383
2384 Default: 1 (enabled)
2385
2386auto_flowlabels - INTEGER
2387 Automatically generate flow labels based on a flow hash of the
2388 packet. This allows intermediate devices, such as routers, to
2389 identify packet flows for mechanisms like Equal Cost Multipath
2390 Routing (see RFC 6438).
2391
2392 = ===========================================================
2393 0 automatic flow labels are completely disabled
2394 1 automatic flow labels are enabled by default, they can be
2395 disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
2396 socket option
2397 2 automatic flow labels are allowed, they may be enabled on a
2398 per socket basis using the IPV6_AUTOFLOWLABEL socket option
2399 3 automatic flow labels are enabled and enforced, they cannot
2400 be disabled by the socket option
2401 = ===========================================================
2402
2403 Default: 1
2404
2405flowlabel_state_ranges - BOOLEAN
2406 Split the flow label number space into two ranges. 0-0x7FFFF is
2407 reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
2408 is reserved for stateless flow labels as described in RFC6437.
2409
2410 Possible values:
2411
2412 - 0 (disabled)
2413 - 1 (enabled)
2414
2415 Default: 1 (enabled)
2416
2417
2418flowlabel_reflect - INTEGER
2419 Control flow label reflection. Needed for Path MTU
2420 Discovery to work with Equal Cost Multipath Routing in anycast
2421 environments. See RFC 7690 and:
2422 https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
2423
2424 This is a bitmask.
2425
2426 - 1: enabled for established flows
2427
2428 Note that this prevents automatic flowlabel changes, as done
2429 in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
2430 and "tcp: Change txhash on every SYN and RTO retransmit"
2431
2432 - 2: enabled for TCP RESET packets (no active listener)
2433 If set, a RST packet sent in response to a SYN packet on a closed
2434 port will reflect the incoming flow label.
2435
2436 - 4: enabled for ICMPv6 echo reply messages.
2437
2438 Default: 0
2439
2440fib_multipath_hash_policy - INTEGER
2441 Controls which hash policy to use for multipath routes.
2442
2443 Default: 0 (Layer 3)
2444
2445 Possible values:
2446
2447 - 0 - Layer 3 (source and destination addresses plus flow label)
2448 - 1 - Layer 4 (standard 5-tuple)
2449 - 2 - Layer 3 or inner Layer 3 if present
2450 - 3 - Custom multipath hash. Fields used for multipath hash calculation
2451 are determined by fib_multipath_hash_fields sysctl
2452
2453fib_multipath_hash_fields - UNSIGNED INTEGER
2454 When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
2455 fields used for multipath hash calculation are determined by this
2456 sysctl.
2457
2458 This value is a bitmask which enables various fields for multipath hash
2459 calculation.
2460
2461 Possible fields are:
2462
2463 ====== ============================
2464 0x0001 Source IP address
2465 0x0002 Destination IP address
2466 0x0004 IP protocol
2467 0x0008 Flow Label
2468 0x0010 Source port
2469 0x0020 Destination port
2470 0x0040 Inner source IP address
2471 0x0080 Inner destination IP address
2472 0x0100 Inner IP protocol
2473 0x0200 Inner Flow Label
2474 0x0400 Inner source port
2475 0x0800 Inner destination port
2476 ====== ============================
2477
2478 Default: 0x0007 (source IP, destination IP and IP protocol)
2479
2480anycast_src_echo_reply - BOOLEAN
2481 Controls the use of anycast addresses as source addresses for ICMPv6
2482 echo reply
2483
2484 Possible values:
2485
2486 - 0 (disabled)
2487 - 1 (enabled)
2488
2489 Default: 0 (disabled)
2490
2491
2492idgen_delay - INTEGER
2493 Controls the delay in seconds after which time to retry
2494 privacy stable address generation if a DAD conflict is
2495 detected.
2496
2497 Default: 1 (as specified in RFC7217)
2498
2499idgen_retries - INTEGER
2500 Controls the number of retries to generate a stable privacy
2501 address if a DAD conflict is detected.
2502
2503 Default: 3 (as specified in RFC7217)
2504
2505mld_qrv - INTEGER
2506 Controls the MLD query robustness variable (see RFC3810 9.1).
2507
2508 Default: 2 (as specified by RFC3810 9.1)
2509
2510 Minimum: 1 (as specified by RFC6636 4.5)
2511
2512max_dst_opts_number - INTEGER
2513 Maximum number of non-padding TLVs allowed in a Destination
2514 options extension header. If this value is less than zero
2515 then unknown options are disallowed and the number of known
2516 TLVs allowed is the absolute value of this number.
2517
2518 Default: 8
2519
2520max_hbh_opts_number - INTEGER
2521 Maximum number of non-padding TLVs allowed in a Hop-by-Hop
2522 options extension header. If this value is less than zero
2523 then unknown options are disallowed and the number of known
2524 TLVs allowed is the absolute value of this number.
2525
2526 Default: 8
2527
2528max_dst_opts_length - INTEGER
2529 Maximum length allowed for a Destination options extension
2530 header.
2531
2532 Default: INT_MAX (unlimited)
2533
2534max_hbh_length - INTEGER
2535 Maximum length allowed for a Hop-by-Hop options extension
2536 header.
2537
2538 Default: INT_MAX (unlimited)
2539
2540skip_notify_on_dev_down - BOOLEAN
2541 Controls whether an RTM_DELROUTE message is generated for routes
2542 removed when a device is taken down or deleted. IPv4 does not
2543 generate this message; IPv6 does by default. Setting this sysctl
2544 to true skips the message, making IPv4 and IPv6 on par in relying
2545 on userspace caches to track link events and evict routes.
2546
2547 Possible values:
2548
2549 - 0 (disabled) - generate the message
2550 - 1 (enabled) - skip generating the message
2551
2552 Default: 0 (disabled)
2553
2554nexthop_compat_mode - BOOLEAN
2555 New nexthop API provides a means for managing nexthops independent of
2556 prefixes. Backwards compatibility with old route format is enabled by
2557 default which means route dumps and notifications contain the new
2558 nexthop attribute but also the full, expanded nexthop definition.
2559 Further, updates or deletes of a nexthop configuration generate route
2560 notifications for each fib entry using the nexthop. Once a system
2561 understands the new API, this sysctl can be disabled to achieve full
2562 performance benefits of the new API by disabling the nexthop expansion
2563 and extraneous notifications.
2564
2565 Note that as a backward-compatible mode, dumping of modern features
2566 might be incomplete or wrong. For example, resilient groups will not be
2567 shown as such, but rather as just a list of next hops. Also weights that
2568 do not fit into 8 bits will show incorrectly.
2569
2570 Default: true (backward compat mode)
2571
2572fib_notify_on_flag_change - INTEGER
2573 Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
2574 RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
2575
2576 After installing a route to the kernel, user space receives an
2577 acknowledgment, which means the route was installed in the kernel,
2578 but not necessarily in hardware.
2579 It is also possible for a route already installed in hardware to change
2580 its action and therefore its flags. For example, a host route that is
2581 trapping packets can be "promoted" to perform decapsulation following
2582 the installation of an IPinIP/VXLAN tunnel.
2583 The notifications will indicate to user-space the state of the route.
2584
2585 Default: 0 (Do not emit notifications.)
2586
2587 Possible values:
2588
2589 - 0 - Do not emit notifications.
2590 - 1 - Emit notifications.
2591 - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
2592
2593ioam6_id - INTEGER
2594 Define the IOAM id of this node. Uses only 24 bits out of 32 in total.
2595
2596 Possible value range:
2597
2598 - Min: 0
2599 - Max: 0xFFFFFF
2600
2601 Default: 0xFFFFFF
2602
2603ioam6_id_wide - LONG INTEGER
2604 Define the wide IOAM id of this node. Uses only 56 bits out of 64 in
2605 total. Can be different from ioam6_id.
2606
2607 Possible value range:
2608
2609 - Min: 0
2610 - Max: 0xFFFFFFFFFFFFFF
2611
2612 Default: 0xFFFFFFFFFFFFFF
2613
2614IPv6 Fragmentation:
2615
2616ip6frag_high_thresh - INTEGER
2617 Maximum memory used to reassemble IPv6 fragments. When
2618 ip6frag_high_thresh bytes of memory is allocated for this purpose,
2619 the fragment handler will toss packets until ip6frag_low_thresh
2620 is reached.
2621
2622ip6frag_low_thresh - INTEGER
2623 See ip6frag_high_thresh
2624
2625ip6frag_time - INTEGER
2626 Time in seconds to keep an IPv6 fragment in memory.
2627
2628``conf/default/*``:
2629 Change the interface-specific default settings.
2630
2631 These settings would be used during creating new interfaces.
2632
2633
2634``conf/all/*``:
2635 Change all the interface-specific settings.
2636
2637 [XXX: Other special features than forwarding?]
2638
2639conf/all/disable_ipv6 - BOOLEAN
2640 Changing this value is same as changing ``conf/default/disable_ipv6``
2641 setting and also all per-interface ``disable_ipv6`` settings to the same
2642 value.
2643
2644 Reading this value does not have any particular meaning. It does not say
2645 whether IPv6 support is enabled or disabled. Returned value can be 1
2646 also in the case when some interface has ``disable_ipv6`` set to 0 and
2647 has configured IPv6 addresses.
2648
2649conf/all/forwarding - BOOLEAN
2650 Enable global IPv6 forwarding between all interfaces.
2651
2652 IPv4 and IPv6 work differently here; the ``force_forwarding`` flag must
2653 be used to control which interfaces may forward packets.
2654
2655 This also sets all interfaces' Host/Router setting
2656 'forwarding' to the specified value. See below for details.
2657
2658 This referred to as global forwarding.
2659
2660proxy_ndp - BOOLEAN
2661 Do proxy ndp.
2662
2663 Possible values:
2664
2665 - 0 (disabled)
2666 - 1 (enabled)
2667
2668 Default: 0 (disabled)
2669
2670force_forwarding - BOOLEAN
2671 Enable forwarding on this interface only -- regardless of the setting on
2672 ``conf/all/forwarding``. When setting ``conf.all.forwarding`` to 0,
2673 the ``force_forwarding`` flag will be reset on all interfaces.
2674
2675fwmark_reflect - BOOLEAN
2676 Controls the fwmark of kernel-generated IPv6 reply packets that are not
2677 associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
2678 If disabled, these packets have a fwmark of zero. If enabled, they have the
2679 fwmark of the packet they are replying to.
2680
2681 Possible values:
2682
2683 - 0 (disabled)
2684 - 1 (enabled)
2685
2686 Default: 0 (disabled)
2687
2688``conf/interface/*``:
2689 Change special settings per interface.
2690
2691 The functional behaviour for certain settings is different
2692 depending on whether local forwarding is enabled or not.
2693
2694accept_ra - INTEGER
2695 Accept Router Advertisements; autoconfigure using them.
2696
2697 It also determines whether or not to transmit Router
2698 Solicitations. If and only if the functional setting is to
2699 accept Router Advertisements, Router Solicitations will be
2700 transmitted.
2701
2702 Possible values are:
2703
2704 == ===========================================================
2705 0 Do not accept Router Advertisements.
2706 1 Accept Router Advertisements if forwarding is disabled.
2707 2 Overrule forwarding behaviour. Accept Router Advertisements
2708 even if forwarding is enabled.
2709 == ===========================================================
2710
2711 Functional default:
2712
2713 - enabled if local forwarding is disabled.
2714 - disabled if local forwarding is enabled.
2715
2716accept_ra_defrtr - BOOLEAN
2717 Learn default router in Router Advertisement.
2718
2719 Functional default:
2720
2721 - enabled if accept_ra is enabled.
2722 - disabled if accept_ra is disabled.
2723
2724ra_defrtr_metric - UNSIGNED INTEGER
2725 Route metric for default route learned in Router Advertisement. This value
2726 will be assigned as metric for the default route learned via IPv6 Router
2727 Advertisement. Takes affect only if accept_ra_defrtr is enabled.
2728
2729 Possible values:
2730 1 to 0xFFFFFFFF
2731
2732 Default: IP6_RT_PRIO_USER i.e. 1024.
2733
2734accept_ra_from_local - BOOLEAN
2735 Accept RA with source-address that is found on local machine
2736 if the RA is otherwise proper and able to be accepted.
2737
2738 Default is to NOT accept these as it may be an un-intended
2739 network loop.
2740
2741 Functional default:
2742
2743 - enabled if accept_ra_from_local is enabled
2744 on a specific interface.
2745 - disabled if accept_ra_from_local is disabled
2746 on a specific interface.
2747
2748accept_ra_min_hop_limit - INTEGER
2749 Minimum hop limit Information in Router Advertisement.
2750
2751 Hop limit Information in Router Advertisement less than this
2752 variable shall be ignored.
2753
2754 Default: 1
2755
2756accept_ra_min_lft - INTEGER
2757 Minimum acceptable lifetime value in Router Advertisement.
2758
2759 RA sections with a lifetime less than this value shall be
2760 ignored. Zero lifetimes stay unaffected.
2761
2762 Default: 0
2763
2764accept_ra_pinfo - BOOLEAN
2765 Learn Prefix Information in Router Advertisement.
2766
2767 Functional default:
2768
2769 - enabled if accept_ra is enabled.
2770 - disabled if accept_ra is disabled.
2771
2772ra_honor_pio_life - BOOLEAN
2773 Whether to use RFC4862 Section 5.5.3e to determine the valid
2774 lifetime of an address matching a prefix sent in a Router
2775 Advertisement Prefix Information Option.
2776
2777 Possible values:
2778
2779 - 0 (disabled) - RFC4862 section 5.5.3e is used to determine
2780 the valid lifetime of the address.
2781 - 1 (enabled) - the PIO valid lifetime will always be honored.
2782
2783 Default: 0 (disabled)
2784
2785ra_honor_pio_pflag - BOOLEAN
2786 The Prefix Information Option P-flag indicates the network can
2787 allocate a unique IPv6 prefix per client using DHCPv6-PD.
2788 This sysctl can be enabled when a userspace DHCPv6-PD client
2789 is running to cause the P-flag to take effect: i.e. the
2790 P-flag suppresses any effects of the A-flag within the same
2791 PIO. For a given PIO, P=1 and A=1 is treated as A=0.
2792
2793 Possible values:
2794
2795 - 0 (disabled) - the P-flag is ignored.
2796 - 1 (enabled) - the P-flag will disable SLAAC autoconfiguration
2797 for the given Prefix Information Option.
2798
2799 Default: 0 (disabled)
2800
2801accept_ra_rt_info_min_plen - INTEGER
2802 Minimum prefix length of Route Information in RA.
2803
2804 Route Information w/ prefix smaller than this variable shall
2805 be ignored.
2806
2807 Functional default:
2808
2809 * 0 if accept_ra_rtr_pref is enabled.
2810 * -1 if accept_ra_rtr_pref is disabled.
2811
2812accept_ra_rt_info_max_plen - INTEGER
2813 Maximum prefix length of Route Information in RA.
2814
2815 Route Information w/ prefix larger than this variable shall
2816 be ignored.
2817
2818 Functional default:
2819
2820 * 0 if accept_ra_rtr_pref is enabled.
2821 * -1 if accept_ra_rtr_pref is disabled.
2822
2823accept_ra_rtr_pref - BOOLEAN
2824 Accept Router Preference in RA.
2825
2826 Functional default:
2827
2828 - enabled if accept_ra is enabled.
2829 - disabled if accept_ra is disabled.
2830
2831accept_ra_mtu - BOOLEAN
2832 Apply the MTU value specified in RA option 5 (RFC4861). If
2833 disabled, the MTU specified in the RA will be ignored.
2834
2835 Functional default:
2836
2837 - enabled if accept_ra is enabled.
2838 - disabled if accept_ra is disabled.
2839
2840accept_redirects - BOOLEAN
2841 Accept Redirects.
2842
2843 Functional default:
2844
2845 - enabled if local forwarding is disabled.
2846 - disabled if local forwarding is enabled.
2847
2848accept_source_route - INTEGER
2849 Accept source routing (routing extension header).
2850
2851 - >= 0: Accept only routing header type 2.
2852 - < 0: Do not accept routing header.
2853
2854 Default: 0
2855
2856autoconf - BOOLEAN
2857 Autoconfigure addresses using Prefix Information in Router
2858 Advertisements.
2859
2860 Functional default:
2861
2862 - enabled if accept_ra_pinfo is enabled.
2863 - disabled if accept_ra_pinfo is disabled.
2864
2865dad_transmits - INTEGER
2866 The amount of Duplicate Address Detection probes to send.
2867
2868 Default: 1
2869
2870forwarding - INTEGER
2871 Configure interface-specific Host/Router behaviour.
2872
2873 .. note::
2874
2875 It is recommended to have the same setting on all
2876 interfaces; mixed router/host scenarios are rather uncommon.
2877
2878 Possible values are:
2879
2880 - 0 Forwarding disabled
2881 - 1 Forwarding enabled
2882
2883 **FALSE (0)**:
2884
2885 By default, Host behaviour is assumed. This means:
2886
2887 1. IsRouter flag is not set in Neighbour Advertisements.
2888 2. If accept_ra is TRUE (default), transmit Router
2889 Solicitations.
2890 3. If accept_ra is TRUE (default), accept Router
2891 Advertisements (and do autoconfiguration).
2892 4. If accept_redirects is TRUE (default), accept Redirects.
2893
2894 **TRUE (1)**:
2895
2896 If local forwarding is enabled, Router behaviour is assumed.
2897 This means exactly the reverse from the above:
2898
2899 1. IsRouter flag is set in Neighbour Advertisements.
2900 2. Router Solicitations are not sent unless accept_ra is 2.
2901 3. Router Advertisements are ignored unless accept_ra is 2.
2902 4. Redirects are ignored.
2903
2904 Default: 0 (disabled) if global forwarding is disabled (default),
2905 otherwise 1 (enabled).
2906
2907hop_limit - INTEGER
2908 Default Hop Limit to set.
2909
2910 Default: 64
2911
2912mtu - INTEGER
2913 Default Maximum Transfer Unit
2914
2915 Default: 1280 (IPv6 required minimum)
2916
2917ip_nonlocal_bind - BOOLEAN
2918 If enabled, allows processes to bind() to non-local IPv6 addresses,
2919 which can be quite useful - but may break some applications.
2920
2921 Possible values:
2922
2923 - 0 (disabled)
2924 - 1 (enabled)
2925
2926 Default: 0 (disabled)
2927
2928router_probe_interval - INTEGER
2929 Minimum interval (in seconds) between Router Probing described
2930 in RFC4191.
2931
2932 Default: 60
2933
2934router_solicitation_delay - INTEGER
2935 Number of seconds to wait after interface is brought up
2936 before sending Router Solicitations.
2937
2938 Default: 1
2939
2940router_solicitation_interval - INTEGER
2941 Number of seconds to wait between Router Solicitations.
2942
2943 Default: 4
2944
2945router_solicitations - INTEGER
2946 Number of Router Solicitations to send until assuming no
2947 routers are present.
2948
2949 Default: 3
2950
2951use_oif_addrs_only - BOOLEAN
2952 When enabled, the candidate source addresses for destinations
2953 routed via this interface are restricted to the set of addresses
2954 configured on this interface (vis. RFC 6724, section 4).
2955
2956 Possible values:
2957
2958 - 0 (disabled)
2959 - 1 (enabled)
2960
2961 Default: 0 (disabled)
2962
2963use_tempaddr - INTEGER
2964 Preference for Privacy Extensions (RFC3041).
2965
2966 * <= 0 : disable Privacy Extensions
2967 * == 1 : enable Privacy Extensions, but prefer public
2968 addresses over temporary addresses.
2969 * > 1 : enable Privacy Extensions and prefer temporary
2970 addresses over public addresses.
2971
2972 Default:
2973
2974 * 0 (for most devices)
2975 * -1 (for point-to-point devices and loopback devices)
2976
2977temp_valid_lft - INTEGER
2978 valid lifetime (in seconds) for temporary addresses. If less than the
2979 minimum required lifetime (typically 5-7 seconds), temporary addresses
2980 will not be created.
2981
2982 Default: 172800 (2 days)
2983
2984temp_prefered_lft - INTEGER
2985 Preferred lifetime (in seconds) for temporary addresses. If
2986 temp_prefered_lft is less than the minimum required lifetime (typically
2987 5-7 seconds), the preferred lifetime is the minimum required. If
2988 temp_prefered_lft is greater than temp_valid_lft, the preferred lifetime
2989 is temp_valid_lft.
2990
2991 Default: 86400 (1 day)
2992
2993keep_addr_on_down - INTEGER
2994 Keep all IPv6 addresses on an interface down event. If set static
2995 global addresses with no expiration time are not flushed.
2996
2997 * >0 : enabled
2998 * 0 : system default
2999 * <0 : disabled
3000
3001 Default: 0 (addresses are removed)
3002
3003max_desync_factor - INTEGER
3004 Maximum value for DESYNC_FACTOR, which is a random value
3005 that ensures that clients don't synchronize with each
3006 other and generate new addresses at exactly the same time.
3007 value is in seconds.
3008
3009 Default: 600
3010
3011regen_min_advance - INTEGER
3012 How far in advance (in seconds), at minimum, to create a new temporary
3013 address before the current one is deprecated. This value is added to
3014 the amount of time that may be required for duplicate address detection
3015 to determine when to create a new address. Linux permits setting this
3016 value to less than the default of 2 seconds, but a value less than 2
3017 does not conform to RFC 8981.
3018
3019 Default: 2
3020
3021regen_max_retry - INTEGER
3022 Number of attempts before give up attempting to generate
3023 valid temporary addresses.
3024
3025 Default: 5
3026
3027max_addresses - INTEGER
3028 Maximum number of autoconfigured addresses per interface. Setting
3029 to zero disables the limitation. It is not recommended to set this
3030 value too large (or to zero) because it would be an easy way to
3031 crash the kernel by allowing too many addresses to be created.
3032
3033 Default: 16
3034
3035disable_ipv6 - BOOLEAN
3036 Disable IPv6 operation. If accept_dad is set to 2, this value
3037 will be dynamically set to TRUE if DAD fails for the link-local
3038 address.
3039
3040 Default: FALSE (enable IPv6 operation)
3041
3042 When this value is changed from 1 to 0 (IPv6 is being enabled),
3043 it will dynamically create a link-local address on the given
3044 interface and start Duplicate Address Detection, if necessary.
3045
3046 When this value is changed from 0 to 1 (IPv6 is being disabled),
3047 it will dynamically delete all addresses and routes on the given
3048 interface. From now on it will not possible to add addresses/routes
3049 to the selected interface.
3050
3051accept_dad - INTEGER
3052 Whether to accept DAD (Duplicate Address Detection).
3053
3054 == ==============================================================
3055 0 Disable DAD
3056 1 Enable DAD (default)
3057 2 Enable DAD, and disable IPv6 operation if MAC-based duplicate
3058 link-local address has been found.
3059 == ==============================================================
3060
3061 DAD operation and mode on a given interface will be selected according
3062 to the maximum value of conf/{all,interface}/accept_dad.
3063
3064force_tllao - BOOLEAN
3065 Enable sending the target link-layer address option even when
3066 responding to a unicast neighbor solicitation.
3067
3068 Default: FALSE
3069
3070 Quoting from RFC 2461, section 4.4, Target link-layer address:
3071
3072 "The option MUST be included for multicast solicitations in order to
3073 avoid infinite Neighbor Solicitation "recursion" when the peer node
3074 does not have a cache entry to return a Neighbor Advertisements
3075 message. When responding to unicast solicitations, the option can be
3076 omitted since the sender of the solicitation has the correct link-
3077 layer address; otherwise it would not have be able to send the unicast
3078 solicitation in the first place. However, including the link-layer
3079 address in this case adds little overhead and eliminates a potential
3080 race condition where the sender deletes the cached link-layer address
3081 prior to receiving a response to a previous solicitation."
3082
3083ndisc_notify - BOOLEAN
3084 Define mode for notification of address and device changes.
3085
3086 Possible values:
3087
3088 - 0 (disabled) - do nothing
3089 - 1 (enabled) - Generate unsolicited neighbour advertisements when device is brought
3090 up or hardware address changes.
3091
3092 Default: 0 (disabled)
3093
3094ndisc_tclass - INTEGER
3095 The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
3096 Discovery (Router Solicitation, Router Advertisement, Neighbor
3097 Solicitation, Neighbor Advertisement, Redirect) messages.
3098 These 8 bits can be interpreted as 6 high order bits holding the DSCP
3099 value and 2 low order bits representing ECN (which you probably want
3100 to leave cleared).
3101
3102 * 0 - (default)
3103
3104ndisc_evict_nocarrier - BOOLEAN
3105 Clears the neighbor discovery table on NOCARRIER events. This option is
3106 important for wireless devices where the neighbor discovery cache should
3107 not be cleared when roaming between access points on the same network.
3108 In most cases this should remain as the default (1).
3109
3110 Possible values:
3111
3112 - 0 (disabled) - Do not clear neighbor discovery cache on NOCARRIER events.
3113 - 1 (enabled) - Clear neighbor discover cache on NOCARRIER events.
3114
3115 Default: 1 (enabled)
3116
3117mldv1_unsolicited_report_interval - INTEGER
3118 The interval in milliseconds in which the next unsolicited
3119 MLDv1 report retransmit will take place.
3120
3121 Default: 10000 (10 seconds)
3122
3123mldv2_unsolicited_report_interval - INTEGER
3124 The interval in milliseconds in which the next unsolicited
3125 MLDv2 report retransmit will take place.
3126
3127 Default: 1000 (1 second)
3128
3129force_mld_version - INTEGER
3130 * 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
3131 * 1 - Enforce to use MLD version 1
3132 * 2 - Enforce to use MLD version 2
3133
3134suppress_frag_ndisc - INTEGER
3135 Control RFC 6980 (Security Implications of IPv6 Fragmentation
3136 with IPv6 Neighbor Discovery) behavior:
3137
3138 * 1 - (default) discard fragmented neighbor discovery packets
3139 * 0 - allow fragmented neighbor discovery packets
3140
3141optimistic_dad - BOOLEAN
3142 Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
3143
3144 Optimistic Duplicate Address Detection for the interface will be enabled
3145 if at least one of conf/{all,interface}/optimistic_dad is set to 1,
3146 it will be disabled otherwise.
3147
3148 Possible values:
3149
3150 - 0 (disabled)
3151 - 1 (enabled)
3152
3153 Default: 0 (disabled)
3154
3155
3156use_optimistic - BOOLEAN
3157 If enabled, do not classify optimistic addresses as deprecated during
3158 source address selection. Preferred addresses will still be chosen
3159 before optimistic addresses, subject to other ranking in the source
3160 address selection algorithm.
3161
3162 This will be enabled if at least one of
3163 conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
3164
3165 Possible values:
3166
3167 - 0 (disabled)
3168 - 1 (enabled)
3169
3170 Default: 0 (disabled)
3171
3172stable_secret - IPv6 address
3173 This IPv6 address will be used as a secret to generate IPv6
3174 addresses for link-local addresses and autoconfigured
3175 ones. All addresses generated after setting this secret will
3176 be stable privacy ones by default. This can be changed via the
3177 addrgenmode ip-link. conf/default/stable_secret is used as the
3178 secret for the namespace, the interface specific ones can
3179 overwrite that. Writes to conf/all/stable_secret are refused.
3180
3181 It is recommended to generate this secret during installation
3182 of a system and keep it stable after that.
3183
3184 By default the stable secret is unset.
3185
3186addr_gen_mode - INTEGER
3187 Defines how link-local and autoconf addresses are generated.
3188
3189 = =================================================================
3190 0 generate address based on EUI64 (default)
3191 1 do no generate a link-local address, use EUI64 for addresses
3192 generated from autoconf
3193 2 generate stable privacy addresses, using the secret from
3194 stable_secret (RFC7217)
3195 3 generate stable privacy addresses, using a random secret if unset
3196 = =================================================================
3197
3198drop_unicast_in_l2_multicast - BOOLEAN
3199 Drop any unicast IPv6 packets that are received in link-layer
3200 multicast (or broadcast) frames.
3201
3202 Possible values:
3203
3204 - 0 (disabled)
3205 - 1 (enabled)
3206
3207 Default: 0 (disabled)
3208
3209drop_unsolicited_na - BOOLEAN
3210 Drop all unsolicited neighbor advertisements, for example if there's
3211 a known good NA proxy on the network and such frames need not be used
3212 (or in the case of 802.11, must not be used to prevent attacks.)
3213
3214 Possible values:
3215
3216 - 0 (disabled)
3217 - 1 (enabled)
3218
3219 Default: 0 (disabled).
3220
3221accept_untracked_na - INTEGER
3222 Define behavior for accepting neighbor advertisements from devices that
3223 are absent in the neighbor cache:
3224
3225 - 0 - (default) Do not accept unsolicited and untracked neighbor
3226 advertisements.
3227
3228 - 1 - Add a new neighbor cache entry in STALE state for routers on
3229 receiving a neighbor advertisement (either solicited or unsolicited)
3230 with target link-layer address option specified if no neighbor entry
3231 is already present for the advertised IPv6 address. Without this knob,
3232 NAs received for untracked addresses (absent in neighbor cache) are
3233 silently ignored.
3234
3235 This is as per router-side behavior documented in RFC9131.
3236
3237 This has lower precedence than drop_unsolicited_na.
3238
3239 This will optimize the return path for the initial off-link
3240 communication that is initiated by a directly connected host, by
3241 ensuring that the first-hop router which turns on this setting doesn't
3242 have to buffer the initial return packets to do neighbor-solicitation.
3243 The prerequisite is that the host is configured to send unsolicited
3244 neighbor advertisements on interface bringup. This setting should be
3245 used in conjunction with the ndisc_notify setting on the host to
3246 satisfy this prerequisite.
3247
3248 - 2 - Extend option (1) to add a new neighbor cache entry only if the
3249 source IP address is in the same subnet as an address configured on
3250 the interface that received the neighbor advertisement.
3251
3252enhanced_dad - BOOLEAN
3253 Include a nonce option in the IPv6 neighbor solicitation messages used for
3254 duplicate address detection per RFC7527. A received DAD NS will only signal
3255 a duplicate address if the nonce is different. This avoids any false
3256 detection of duplicates due to loopback of the NS messages that we send.
3257 The nonce option will be sent on an interface unless both of
3258 conf/{all,interface}/enhanced_dad are set to FALSE.
3259
3260 Possible values:
3261
3262 - 0 (disabled)
3263 - 1 (enabled)
3264
3265 Default: 1 (enabled)
3266
3267``icmp/*``:
3268===========
3269
3270ratelimit - INTEGER
3271 Limit the maximal rates for sending ICMPv6 messages to a particular
3272 peer.
3273
3274 0 to disable any limiting,
3275 otherwise the space between responses in milliseconds.
3276
3277 Default: 100
3278
3279ratemask - list of comma separated ranges
3280 For ICMPv6 message types matching the ranges in the ratemask, limit
3281 the sending of the message according to ratelimit parameter.
3282
3283 The format used for both input and output is a comma separated
3284 list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
3285 129). Writing to the file will clear all previous ranges of ICMPv6
3286 message types and update the current list with the input.
3287
3288 Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
3289 for numerical values of ICMPv6 message types, e.g. echo request is 128
3290 and echo reply is 129.
3291
3292 Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
3293
3294echo_ignore_all - BOOLEAN
3295 If enabled, then the kernel will ignore all ICMP ECHO
3296 requests sent to it over the IPv6 protocol.
3297
3298 Possible values:
3299
3300 - 0 (disabled)
3301 - 1 (enabled)
3302
3303 Default: 0 (disabled)
3304
3305echo_ignore_multicast - BOOLEAN
3306 If enabled, then the kernel will ignore all ICMP ECHO
3307 requests sent to it over the IPv6 protocol via multicast.
3308
3309 Possible values:
3310
3311 - 0 (disabled)
3312 - 1 (enabled)
3313
3314 Default: 0 (disabled)
3315
3316echo_ignore_anycast - BOOLEAN
3317 If enabled, then the kernel will ignore all ICMP ECHO
3318 requests sent to it over the IPv6 protocol destined to anycast address.
3319
3320 Possible values:
3321
3322 - 0 (disabled)
3323 - 1 (enabled)
3324
3325 Default: 0 (disabled)
3326
3327error_anycast_as_unicast - BOOLEAN
3328 If enabled, then the kernel will respond with ICMP Errors
3329 resulting from requests sent to it over the IPv6 protocol destined
3330 to anycast address essentially treating anycast as unicast.
3331
3332 Possible values:
3333
3334 - 0 (disabled)
3335 - 1 (enabled)
3336
3337 Default: 0 (disabled)
3338
3339errors_extension_mask - UNSIGNED INTEGER
3340 Bitmask of ICMP extensions to append to ICMPv6 error messages
3341 ("Destination Unreachable" and "Time Exceeded"). The original datagram
3342 is trimmed / padded to 128 bytes in order to be compatible with
3343 applications that do not comply with RFC 4884.
3344
3345 Possible extensions are:
3346
3347 ==== ==============================================================
3348 0x01 Incoming IP interface information according to RFC 5837.
3349 Extension will include the index, IPv6 address (if present),
3350 name and MTU of the IP interface that received the datagram
3351 which elicited the ICMP error.
3352 ==== ==============================================================
3353
3354 Default: 0x00 (no extensions)
3355
3356xfrm6_gc_thresh - INTEGER
3357 (Obsolete since linux-4.14)
3358 The threshold at which we will start garbage collecting for IPv6
3359 destination cache entries. At twice this value the system will
3360 refuse new allocations.
3361
3362
3363IPv6 Update by:
3364Pekka Savola <pekkas@netcore.fi>
3365YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
3366
3367
3368/proc/sys/net/bridge/* Variables:
3369=================================
3370
3371bridge-nf-call-arptables - BOOLEAN
3372
3373 Possible values:
3374
3375 - 0 (disabled) - disable this.
3376 - 1 (enabled) - pass bridged ARP traffic to arptables' FORWARD chain.
3377
3378 Default: 1 (enabled)
3379
3380bridge-nf-call-iptables - BOOLEAN
3381
3382 Possible values:
3383
3384 - 0 (disabled) - disable this.
3385 - 1 (enabled) - pass bridged IPv4 traffic to iptables' chains.
3386
3387 Default: 1 (enabled)
3388
3389bridge-nf-call-ip6tables - BOOLEAN
3390
3391 Possible values:
3392
3393 - 0 (disabled) - disable this.
3394 - 1 (enabled) - pass bridged IPv6 traffic to ip6tables' chains.
3395
3396 Default: 1 (enabled)
3397
3398bridge-nf-filter-vlan-tagged - BOOLEAN
3399
3400 Possible values:
3401
3402 - 0 (disabled) - disable this.
3403 - 1 (enabled) - pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables
3404
3405 Default: 0 (disabled)
3406
3407bridge-nf-filter-pppoe-tagged - BOOLEAN
3408
3409 Possible values:
3410
3411 - 0 (disabled) - disable this.
3412 - 1 (enabled) - pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
3413
3414 Default: 0 (disabled)
3415
3416bridge-nf-pass-vlan-input-dev - BOOLEAN
3417 - 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
3418 interface on the bridge and set the netfilter input device to the
3419 vlan. This allows use of e.g. "iptables -i br0.1" and makes the
3420 REDIRECT target work with vlan-on-top-of-bridge interfaces. When no
3421 matching vlan interface is found, or this switch is off, the input
3422 device is set to the bridge interface.
3423
3424 - 0: disable bridge netfilter vlan interface lookup.
3425
3426 Default: 0
3427
3428``proc/sys/net/sctp/*`` Variables:
3429==================================
3430
3431addip_enable - BOOLEAN
3432 Enable or disable extension of Dynamic Address Reconfiguration
3433 (ADD-IP) functionality specified in RFC5061. This extension provides
3434 the ability to dynamically add and remove new addresses for the SCTP
3435 associations.
3436
3437 Possible values:
3438
3439 - 0 (disabled) - disable extension.
3440 - 1 (enabled) - enable extension
3441
3442 Default: 0 (disabled)
3443
3444pf_enable - INTEGER
3445 Enable or disable pf (pf is short for potentially failed) state. A value
3446 of pf_retrans > path_max_retrans also disables pf state. That is, one of
3447 both pf_enable and pf_retrans > path_max_retrans can disable pf state.
3448 Since pf_retrans and path_max_retrans can be changed by userspace
3449 application, sometimes user expects to disable pf state by the value of
3450 pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
3451 or path_max_retrans is changed by the user application, this pf state is
3452 enabled. As such, it is necessary to add this to dynamically enable
3453 and disable pf state. See:
3454 https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
3455 details.
3456
3457 Possible values:
3458
3459 - 1: Enable pf.
3460 - 0: Disable pf.
3461
3462 Default: 1
3463
3464pf_expose - INTEGER
3465 Unset or enable/disable pf (pf is short for potentially failed) state
3466 exposure. Applications can control the exposure of the PF path state
3467 in the SCTP_PEER_ADDR_CHANGE event and access of SCTP_PF-state
3468 transport info via SCTP_GET_PEER_ADDR_INFO sockopt.
3469
3470 Possible values:
3471
3472 - 0: Unset pf state exposure (compatible with old applications). No
3473 event will be sent but the transport info can be queried.
3474 - 1: Disable pf state exposure. No event will be sent and trying to
3475 obtain transport info will return -EACCESS.
3476 - 2: Enable pf state exposure. The event will be sent for a transport
3477 becoming SCTP_PF state and transport info can be obtained.
3478
3479 Default: 0
3480
3481addip_noauth_enable - BOOLEAN
3482 Dynamic Address Reconfiguration (ADD-IP) requires the use of
3483 authentication to protect the operations of adding or removing new
3484 addresses. This requirement is mandated so that unauthorized hosts
3485 would not be able to hijack associations. However, older
3486 implementations may not have implemented this requirement while
3487 allowing the ADD-IP extension. For reasons of interoperability,
3488 we provide this variable to control the enforcement of the
3489 authentication requirement.
3490
3491 == ===============================================================
3492 1 Allow ADD-IP extension to be used without authentication. This
3493 should only be set in a closed environment for interoperability
3494 with older implementations.
3495
3496 0 Enforce the authentication requirement
3497 == ===============================================================
3498
3499 Default: 0
3500
3501auth_enable - BOOLEAN
3502 Enable or disable Authenticated Chunks extension. This extension
3503 provides the ability to send and receive authenticated chunks and is
3504 required for secure operation of Dynamic Address Reconfiguration
3505 (ADD-IP) extension.
3506
3507 Possible values:
3508
3509 - 0 (disabled) - disable extension.
3510 - 1 (enabled) - enable extension
3511
3512 Default: 0 (disabled)
3513
3514prsctp_enable - BOOLEAN
3515 Enable or disable the Partial Reliability extension (RFC3758) which
3516 is used to notify peers that a given DATA should no longer be expected.
3517
3518 Possible values:
3519
3520 - 0 (disabled) - disable extension.
3521 - 1 (enabled) - enable extension
3522
3523 Default: 1 (enabled)
3524
3525max_burst - INTEGER
3526 The limit of the number of new packets that can be initially sent. It
3527 controls how bursty the generated traffic can be.
3528
3529 Default: 4
3530
3531association_max_retrans - INTEGER
3532 Set the maximum number for retransmissions that an association can
3533 attempt deciding that the remote end is unreachable. If this value
3534 is exceeded, the association is terminated.
3535
3536 Default: 10
3537
3538max_init_retransmits - INTEGER
3539 The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
3540 that an association will attempt before declaring the destination
3541 unreachable and terminating.
3542
3543 Default: 8
3544
3545path_max_retrans - INTEGER
3546 The maximum number of retransmissions that will be attempted on a given
3547 path. Once this threshold is exceeded, the path is considered
3548 unreachable, and new traffic will use a different path when the
3549 association is multihomed.
3550
3551 Default: 5
3552
3553pf_retrans - INTEGER
3554 The number of retransmissions that will be attempted on a given path
3555 before traffic is redirected to an alternate transport (should one
3556 exist). Note this is distinct from path_max_retrans, as a path that
3557 passes the pf_retrans threshold can still be used. Its only
3558 deprioritized when a transmission path is selected by the stack. This
3559 setting is primarily used to enable fast failover mechanisms without
3560 having to reduce path_max_retrans to a very low value. See:
3561 http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3562 for details. Note also that a value of pf_retrans > path_max_retrans
3563 disables this feature. Since both pf_retrans and path_max_retrans can
3564 be changed by userspace application, a variable pf_enable is used to
3565 disable pf state.
3566
3567 Default: 0
3568
3569ps_retrans - INTEGER
3570 Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
3571 from section-5 "Primary Path Switchover" in rfc7829. The primary path
3572 will be changed to another active path when the path error counter on
3573 the old primary path exceeds PSMR, so that "the SCTP sender is allowed
3574 to continue data transmission on a new working path even when the old
3575 primary destination address becomes active again". Note this feature
3576 is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
3577 and its value can't be less than 'pf_retrans' when changing by sysctl.
3578
3579 Default: 0xffff
3580
3581rto_initial - INTEGER
3582 The initial round trip timeout value in milliseconds that will be used
3583 in calculating round trip times. This is the initial time interval
3584 for retransmissions.
3585
3586 Default: 3000
3587
3588rto_max - INTEGER
3589 The maximum value (in milliseconds) of the round trip timeout. This
3590 is the largest time interval that can elapse between retransmissions.
3591
3592 Default: 60000
3593
3594rto_min - INTEGER
3595 The minimum value (in milliseconds) of the round trip timeout. This
3596 is the smallest time interval the can elapse between retransmissions.
3597
3598 Default: 1000
3599
3600hb_interval - INTEGER
3601 The interval (in milliseconds) between HEARTBEAT chunks. These chunks
3602 are sent at the specified interval on idle paths to probe the state of
3603 a given path between 2 associations.
3604
3605 Default: 30000
3606
3607sack_timeout - INTEGER
3608 The amount of time (in milliseconds) that the implementation will wait
3609 to send a SACK.
3610
3611 Default: 200
3612
3613valid_cookie_life - INTEGER
3614 The default lifetime of the SCTP cookie (in milliseconds). The cookie
3615 is used during association establishment.
3616
3617 Default: 60000
3618
3619cookie_preserve_enable - BOOLEAN
3620 Enable or disable the ability to extend the lifetime of the SCTP cookie
3621 that is used during the establishment phase of SCTP association
3622
3623 Possible values:
3624
3625 - 0 (disabled) - disable.
3626 - 1 (enabled) - enable cookie lifetime extension.
3627
3628 Default: 1 (enabled)
3629
3630cookie_hmac_alg - STRING
3631 Select the hmac algorithm used when generating the cookie value sent by
3632 a listening sctp socket to a connecting client in the INIT-ACK chunk.
3633 Valid values are:
3634
3635 * sha256
3636 * none
3637
3638 Default: sha256
3639
3640rcvbuf_policy - INTEGER
3641 Determines if the receive buffer is attributed to the socket or to
3642 association. SCTP supports the capability to create multiple
3643 associations on a single socket. When using this capability, it is
3644 possible that a single stalled association that's buffering a lot
3645 of data may block other associations from delivering their data by
3646 consuming all of the receive buffer space. To work around this,
3647 the rcvbuf_policy could be set to attribute the receiver buffer space
3648 to each association instead of the socket. This prevents the described
3649 blocking.
3650
3651 - 1: rcvbuf space is per association
3652 - 0: rcvbuf space is per socket
3653
3654 Default: 0
3655
3656sndbuf_policy - INTEGER
3657 Similar to rcvbuf_policy above, this applies to send buffer space.
3658
3659 - 1: Send buffer is tracked per association
3660 - 0: Send buffer is tracked per socket.
3661
3662 Default: 0
3663
3664sctp_mem - vector of 3 INTEGERs: min, pressure, max
3665 Number of pages allowed for queueing by all SCTP sockets.
3666
3667 * min: Below this number of pages SCTP is not bothered about its
3668 memory usage. When amount of memory allocated by SCTP exceeds
3669 this number, SCTP starts to moderate memory usage.
3670 * pressure: This value was introduced to follow format of tcp_mem.
3671 * max: Maximum number of allowed pages.
3672
3673 Default is calculated at boot time from amount of available memory.
3674
3675sctp_rmem - vector of 3 INTEGERs: min, default, max
3676 Only the first value ("min") is used, "default" and "max" are
3677 ignored.
3678
3679 * min: Minimal size of receive buffer used by SCTP socket.
3680 It is guaranteed to each SCTP socket (but not association) even
3681 under moderate memory pressure.
3682
3683 Default: 4K
3684
3685sctp_wmem - vector of 3 INTEGERs: min, default, max
3686 Only the first value ("min") is used, "default" and "max" are
3687 ignored.
3688
3689 * min: Minimum size of send buffer that can be used by SCTP sockets.
3690 It is guaranteed to each SCTP socket (but not association) even
3691 under moderate memory pressure.
3692
3693 Default: 4K
3694
3695addr_scope_policy - INTEGER
3696 Control IPv4 address scoping (see
3697 https://datatracker.ietf.org/doc/draft-stewart-tsvwg-sctp-ipv4/00/
3698 for details).
3699
3700 - 0 - Disable IPv4 address scoping
3701 - 1 - Enable IPv4 address scoping
3702 - 2 - Follow draft but allow IPv4 private addresses
3703 - 3 - Follow draft but allow IPv4 link local addresses
3704
3705 Default: 1
3706
3707udp_port - INTEGER
3708 The listening port for the local UDP tunneling sock. Normally it's
3709 using the IANA-assigned UDP port number 9899 (sctp-tunneling).
3710
3711 This UDP sock is used for processing the incoming UDP-encapsulated
3712 SCTP packets (from RFC6951), and shared by all applications in the
3713 same net namespace. This UDP sock will be closed when the value is
3714 set to 0.
3715
3716 The value will also be used to set the src port of the UDP header
3717 for the outgoing UDP-encapsulated SCTP packets. For the dest port,
3718 please refer to 'encap_port' below.
3719
3720 Default: 0
3721
3722encap_port - INTEGER
3723 The default remote UDP encapsulation port.
3724
3725 This value is used to set the dest port of the UDP header for the
3726 outgoing UDP-encapsulated SCTP packets by default. Users can also
3727 change the value for each sock/asoc/transport by using setsockopt.
3728 For further information, please refer to RFC6951.
3729
3730 Note that when connecting to a remote server, the client should set
3731 this to the port that the UDP tunneling sock on the peer server is
3732 listening to and the local UDP tunneling sock on the client also
3733 must be started. On the server, it would get the encap_port from
3734 the incoming packet's source port.
3735
3736 Default: 0
3737
3738plpmtud_probe_interval - INTEGER
3739 The time interval (in milliseconds) for the PLPMTUD probe timer,
3740 which is configured to expire after this period to receive an
3741 acknowledgment to a probe packet. This is also the time interval
3742 between the probes for the current pmtu when the probe search
3743 is done.
3744
3745 PLPMTUD will be disabled when 0 is set, and other values for it
3746 must be >= 5000.
3747
3748 Default: 0
3749
3750reconf_enable - BOOLEAN
3751 Enable or disable extension of Stream Reconfiguration functionality
3752 specified in RFC6525. This extension provides the ability to "reset"
3753 a stream, and it includes the Parameters of "Outgoing/Incoming SSN
3754 Reset", "SSN/TSN Reset" and "Add Outgoing/Incoming Streams".
3755
3756 Possible values:
3757
3758 - 0 (disabled) - Disable extension.
3759 - 1 (enabled) - Enable extension.
3760
3761 Default: 0 (disabled)
3762
3763intl_enable - BOOLEAN
3764 Enable or disable extension of User Message Interleaving functionality
3765 specified in RFC8260. This extension allows the interleaving of user
3766 messages sent on different streams. With this feature enabled, I-DATA
3767 chunk will replace DATA chunk to carry user messages if also supported
3768 by the peer. Note that to use this feature, one needs to set this option
3769 to 1 and also needs to set socket options SCTP_FRAGMENT_INTERLEAVE to 2
3770 and SCTP_INTERLEAVING_SUPPORTED to 1.
3771
3772 Possible values:
3773
3774 - 0 (disabled) - Disable extension.
3775 - 1 (enabled) - Enable extension.
3776
3777 Default: 0 (disabled)
3778
3779ecn_enable - BOOLEAN
3780 Control use of Explicit Congestion Notification (ECN) by SCTP.
3781 Like in TCP, ECN is used only when both ends of the SCTP connection
3782 indicate support for it. This feature is useful in avoiding losses
3783 due to congestion by allowing supporting routers to signal congestion
3784 before having to drop packets.
3785
3786 Possible values:
3787
3788 - 0 (disabled) - Disable ecn.
3789 - 1 (enabled) - Enable ecn.
3790
3791 Default: 1 (enabled)
3792
3793l3mdev_accept - BOOLEAN
3794 Enabling this option allows a "global" bound socket to work
3795 across L3 master domains (e.g., VRFs) with packets capable of
3796 being received regardless of the L3 domain in which they
3797 originated. Only valid when the kernel was compiled with
3798 CONFIG_NET_L3_MASTER_DEV.
3799
3800 Possible values:
3801
3802 - 0 (disabled)
3803 - 1 (enabled)
3804
3805 Default: 1 (enabled)
3806
3807
3808``/proc/sys/net/core/*``
3809========================
3810
3811 Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
3812
3813
3814``/proc/sys/net/unix/*``
3815========================
3816
3817max_dgram_qlen - INTEGER
3818 The maximum length of dgram socket receive queue
3819
3820 Default: 10
3821