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	- 0 - disabled (default)
  12	- not 0 - enabled
  13
  14	Forward Packets between interfaces.
  15
  16	This variable is special, its change resets all configuration
  17	parameters to their default state (RFC1122 for hosts, RFC1812
  18	for routers)
  19
  20ip_default_ttl - INTEGER
  21	Default value of TTL field (Time To Live) for outgoing (but not
  22	forwarded) IP packets. Should be between 1 and 255 inclusive.
  23	Default: 64 (as recommended by RFC1700)
  24
  25ip_no_pmtu_disc - INTEGER
  26	Disable Path MTU Discovery. If enabled in mode 1 and a
  27	fragmentation-required ICMP is received, the PMTU to this
  28	destination will be set to the smallest of the old MTU to
  29	this destination and min_pmtu (see below). You will need
  30	to raise min_pmtu to the smallest interface MTU on your system
  31	manually if you want to avoid locally generated fragments.
  32
  33	In mode 2 incoming Path MTU Discovery messages will be
  34	discarded. Outgoing frames are handled the same as in mode 1,
  35	implicitly setting IP_PMTUDISC_DONT on every created socket.
  36
  37	Mode 3 is a hardened pmtu discover mode. The kernel will only
  38	accept fragmentation-needed errors if the underlying protocol
  39	can verify them besides a plain socket lookup. Current
  40	protocols for which pmtu events will be honored are TCP, SCTP
  41	and DCCP as they verify e.g. the sequence number or the
  42	association. This mode should not be enabled globally but is
  43	only intended to secure e.g. name servers in namespaces where
  44	TCP path mtu must still work but path MTU information of other
  45	protocols should be discarded. If enabled globally this mode
  46	could break other protocols.
  47
  48	Possible values: 0-3
  49
  50	Default: FALSE
  51
  52min_pmtu - INTEGER
  53	default 552 - minimum Path MTU. Unless this is changed manually,
  54	each cached pmtu will never be lower than this setting.
  55
  56ip_forward_use_pmtu - BOOLEAN
  57	By default we don't trust protocol path MTUs while forwarding
  58	because they could be easily forged and can lead to unwanted
  59	fragmentation by the router.
  60	You only need to enable this if you have user-space software
  61	which tries to discover path mtus by itself and depends on the
  62	kernel honoring this information. This is normally not the
  63	case.
  64
  65	Default: 0 (disabled)
  66
  67	Possible values:
  68
  69	- 0 - disabled
  70	- 1 - enabled
  71
  72fwmark_reflect - BOOLEAN
  73	Controls the fwmark of kernel-generated IPv4 reply packets that are not
  74	associated with a socket for example, TCP RSTs or ICMP echo replies).
  75	If unset, these packets have a fwmark of zero. If set, they have the
  76	fwmark of the packet they are replying to.
  77
  78	Default: 0
  79
  80fib_multipath_use_neigh - BOOLEAN
  81	Use status of existing neighbor entry when determining nexthop for
  82	multipath routes. If disabled, neighbor information is not used and
  83	packets could be directed to a failed nexthop. Only valid for kernels
  84	built with CONFIG_IP_ROUTE_MULTIPATH enabled.
  85
  86	Default: 0 (disabled)
  87
  88	Possible values:
  89
  90	- 0 - disabled
  91	- 1 - enabled
  92
  93fib_multipath_hash_policy - INTEGER
  94	Controls which hash policy to use for multipath routes. Only valid
  95	for kernels built with CONFIG_IP_ROUTE_MULTIPATH enabled.
  96
  97	Default: 0 (Layer 3)
  98
  99	Possible values:
 100
 101	- 0 - Layer 3
 102	- 1 - Layer 4
 103	- 2 - Layer 3 or inner Layer 3 if present
 104	- 3 - Custom multipath hash. Fields used for multipath hash calculation
 105	  are determined by fib_multipath_hash_fields sysctl
 106
 107fib_multipath_hash_fields - UNSIGNED INTEGER
 108	When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
 109	fields used for multipath hash calculation are determined by this
 110	sysctl.
 111
 112	This value is a bitmask which enables various fields for multipath hash
 113	calculation.
 114
 115	Possible fields are:
 116
 117	====== ============================
 118	0x0001 Source IP address
 119	0x0002 Destination IP address
 120	0x0004 IP protocol
 121	0x0008 Unused (Flow Label)
 122	0x0010 Source port
 123	0x0020 Destination port
 124	0x0040 Inner source IP address
 125	0x0080 Inner destination IP address
 126	0x0100 Inner IP protocol
 127	0x0200 Inner Flow Label
 128	0x0400 Inner source port
 129	0x0800 Inner destination port
 130	====== ============================
 131
 132	Default: 0x0007 (source IP, destination IP and IP protocol)
 133
 134fib_multipath_hash_seed - UNSIGNED INTEGER
 135	The seed value used when calculating hash for multipath routes. Applies
 136	to both IPv4 and IPv6 datapath. Only present for kernels built with
 137	CONFIG_IP_ROUTE_MULTIPATH enabled.
 138
 139	When set to 0, the seed value used for multipath routing defaults to an
 140	internal random-generated one.
 141
 142	The actual hashing algorithm is not specified -- there is no guarantee
 143	that a next hop distribution effected by a given seed will keep stable
 144	across kernel versions.
 145
 146	Default: 0 (random)
 147
 148fib_sync_mem - UNSIGNED INTEGER
 149	Amount of dirty memory from fib entries that can be backlogged before
 150	synchronize_rcu is forced.
 151
 152	Default: 512kB   Minimum: 64kB   Maximum: 64MB
 153
 154ip_forward_update_priority - INTEGER
 155	Whether to update SKB priority from "TOS" field in IPv4 header after it
 156	is forwarded. The new SKB priority is mapped from TOS field value
 157	according to an rt_tos2priority table (see e.g. man tc-prio).
 158
 159	Default: 1 (Update priority.)
 160
 161	Possible values:
 162
 163	- 0 - Do not update priority.
 164	- 1 - Update priority.
 165
 166route/max_size - INTEGER
 167	Maximum number of routes allowed in the kernel.  Increase
 168	this when using large numbers of interfaces and/or routes.
 169
 170	From linux kernel 3.6 onwards, this is deprecated for ipv4
 171	as route cache is no longer used.
 172
 173	From linux kernel 6.3 onwards, this is deprecated for ipv6
 174	as garbage collection manages cached route entries.
 175
 176neigh/default/gc_thresh1 - INTEGER
 177	Minimum number of entries to keep.  Garbage collector will not
 178	purge entries if there are fewer than this number.
 179
 180	Default: 128
 181
 182neigh/default/gc_thresh2 - INTEGER
 183	Threshold when garbage collector becomes more aggressive about
 184	purging entries. Entries older than 5 seconds will be cleared
 185	when over this number.
 186
 187	Default: 512
 188
 189neigh/default/gc_thresh3 - INTEGER
 190	Maximum number of non-PERMANENT neighbor entries allowed.  Increase
 191	this when using large numbers of interfaces and when communicating
 192	with large numbers of directly-connected peers.
 193
 194	Default: 1024
 195
 196neigh/default/unres_qlen_bytes - INTEGER
 197	The maximum number of bytes which may be used by packets
 198	queued for each	unresolved address by other network layers.
 199	(added in linux 3.3)
 200
 201	Setting negative value is meaningless and will return error.
 202
 203	Default: SK_WMEM_MAX, (same as net.core.wmem_default).
 204
 205		Exact value depends on architecture and kernel options,
 206		but should be enough to allow queuing 256 packets
 207		of medium size.
 208
 209neigh/default/unres_qlen - INTEGER
 210	The maximum number of packets which may be queued for each
 211	unresolved address by other network layers.
 212
 213	(deprecated in linux 3.3) : use unres_qlen_bytes instead.
 214
 215	Prior to linux 3.3, the default value is 3 which may cause
 216	unexpected packet loss. The current default value is calculated
 217	according to default value of unres_qlen_bytes and true size of
 218	packet.
 219
 220	Default: 101
 221
 222neigh/default/interval_probe_time_ms - INTEGER
 223	The probe interval for neighbor entries with NTF_MANAGED flag,
 224	the min value is 1.
 225
 226	Default: 5000
 227
 228mtu_expires - INTEGER
 229	Time, in seconds, that cached PMTU information is kept.
 230
 231min_adv_mss - INTEGER
 232	The advertised MSS depends on the first hop route MTU, but will
 233	never be lower than this setting.
 234
 235fib_notify_on_flag_change - INTEGER
 236        Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
 237        RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
 238
 239        After installing a route to the kernel, user space receives an
 240        acknowledgment, which means the route was installed in the kernel,
 241        but not necessarily in hardware.
 242        It is also possible for a route already installed in hardware to change
 243        its action and therefore its flags. For example, a host route that is
 244        trapping packets can be "promoted" to perform decapsulation following
 245        the installation of an IPinIP/VXLAN tunnel.
 246        The notifications will indicate to user-space the state of the route.
 247
 248        Default: 0 (Do not emit notifications.)
 249
 250        Possible values:
 251
 252        - 0 - Do not emit notifications.
 253        - 1 - Emit notifications.
 254        - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
 255
 256IP Fragmentation:
 257
 258ipfrag_high_thresh - LONG INTEGER
 259	Maximum memory used to reassemble IP fragments.
 260
 261ipfrag_low_thresh - LONG INTEGER
 262	(Obsolete since linux-4.17)
 263	Maximum memory used to reassemble IP fragments before the kernel
 264	begins to remove incomplete fragment queues to free up resources.
 265	The kernel still accepts new fragments for defragmentation.
 266
 267ipfrag_time - INTEGER
 268	Time in seconds to keep an IP fragment in memory.
 269
 270ipfrag_max_dist - INTEGER
 271	ipfrag_max_dist is a non-negative integer value which defines the
 272	maximum "disorder" which is allowed among fragments which share a
 273	common IP source address. Note that reordering of packets is
 274	not unusual, but if a large number of fragments arrive from a source
 275	IP address while a particular fragment queue remains incomplete, it
 276	probably indicates that one or more fragments belonging to that queue
 277	have been lost. When ipfrag_max_dist is positive, an additional check
 278	is done on fragments before they are added to a reassembly queue - if
 279	ipfrag_max_dist (or more) fragments have arrived from a particular IP
 280	address between additions to any IP fragment queue using that source
 281	address, it's presumed that one or more fragments in the queue are
 282	lost. The existing fragment queue will be dropped, and a new one
 283	started. An ipfrag_max_dist value of zero disables this check.
 284
 285	Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
 286	result in unnecessarily dropping fragment queues when normal
 287	reordering of packets occurs, which could lead to poor application
 288	performance. Using a very large value, e.g. 50000, increases the
 289	likelihood of incorrectly reassembling IP fragments that originate
 290	from different IP datagrams, which could result in data corruption.
 291	Default: 64
 292
 293bc_forwarding - INTEGER
 294	bc_forwarding enables the feature described in rfc1812#section-5.3.5.2
 295	and rfc2644. It allows the router to forward directed broadcast.
 296	To enable this feature, the 'all' entry and the input interface entry
 297	should be set to 1.
 298	Default: 0
 299
 300INET peer storage
 301=================
 302
 303inet_peer_threshold - INTEGER
 304	The approximate size of the storage.  Starting from this threshold
 305	entries will be thrown aggressively.  This threshold also determines
 306	entries' time-to-live and time intervals between garbage collection
 307	passes.  More entries, less time-to-live, less GC interval.
 308
 309inet_peer_minttl - INTEGER
 310	Minimum time-to-live of entries.  Should be enough to cover fragment
 311	time-to-live on the reassembling side.  This minimum time-to-live  is
 312	guaranteed if the pool size is less than inet_peer_threshold.
 313	Measured in seconds.
 314
 315inet_peer_maxttl - INTEGER
 316	Maximum time-to-live of entries.  Unused entries will expire after
 317	this period of time if there is no memory pressure on the pool (i.e.
 318	when the number of entries in the pool is very small).
 319	Measured in seconds.
 320
 321TCP variables
 322=============
 323
 324somaxconn - INTEGER
 325	Limit of socket listen() backlog, known in userspace as SOMAXCONN.
 326	Defaults to 4096. (Was 128 before linux-5.4)
 327	See also tcp_max_syn_backlog for additional tuning for TCP sockets.
 328
 329tcp_abort_on_overflow - BOOLEAN
 330	If listening service is too slow to accept new connections,
 331	reset them. Default state is FALSE. It means that if overflow
 332	occurred due to a burst, connection will recover. Enable this
 333	option _only_ if you are really sure that listening daemon
 334	cannot be tuned to accept connections faster. Enabling this
 335	option can harm clients of your server.
 336
 337tcp_adv_win_scale - INTEGER
 338	Obsolete since linux-6.6
 339	Count buffering overhead as bytes/2^tcp_adv_win_scale
 340	(if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
 341	if it is <= 0.
 342
 343	Possible values are [-31, 31], inclusive.
 344
 345	Default: 1
 346
 347tcp_allowed_congestion_control - STRING
 348	Show/set the congestion control choices available to non-privileged
 349	processes. The list is a subset of those listed in
 350	tcp_available_congestion_control.
 351
 352	Default is "reno" and the default setting (tcp_congestion_control).
 353
 354tcp_app_win - INTEGER
 355	Reserve max(window/2^tcp_app_win, mss) of window for application
 356	buffer. Value 0 is special, it means that nothing is reserved.
 357
 358	Possible values are [0, 31], inclusive.
 359
 360	Default: 31
 361
 362tcp_autocorking - BOOLEAN
 363	Enable TCP auto corking :
 364	When applications do consecutive small write()/sendmsg() system calls,
 365	we try to coalesce these small writes as much as possible, to lower
 366	total amount of sent packets. This is done if at least one prior
 367	packet for the flow is waiting in Qdisc queues or device transmit
 368	queue. Applications can still use TCP_CORK for optimal behavior
 369	when they know how/when to uncork their sockets.
 370
 371	Default : 1
 372
 373tcp_available_congestion_control - STRING
 374	Shows the available congestion control choices that are registered.
 375	More congestion control algorithms may be available as modules,
 376	but not loaded.
 377
 378tcp_base_mss - INTEGER
 379	The initial value of search_low to be used by the packetization layer
 380	Path MTU discovery (MTU probing).  If MTU probing is enabled,
 381	this is the initial MSS used by the connection.
 382
 383tcp_mtu_probe_floor - INTEGER
 384	If MTU probing is enabled this caps the minimum MSS used for search_low
 385	for the connection.
 386
 387	Default : 48
 388
 389tcp_min_snd_mss - INTEGER
 390	TCP SYN and SYNACK messages usually advertise an ADVMSS option,
 391	as described in RFC 1122 and RFC 6691.
 392
 393	If this ADVMSS option is smaller than tcp_min_snd_mss,
 394	it is silently capped to tcp_min_snd_mss.
 395
 396	Default : 48 (at least 8 bytes of payload per segment)
 397
 398tcp_congestion_control - STRING
 399	Set the congestion control algorithm to be used for new
 400	connections. The algorithm "reno" is always available, but
 401	additional choices may be available based on kernel configuration.
 402	Default is set as part of kernel configuration.
 403	For passive connections, the listener congestion control choice
 404	is inherited.
 405
 406	[see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
 407
 408tcp_dsack - BOOLEAN
 409	Allows TCP to send "duplicate" SACKs.
 410
 411tcp_early_retrans - INTEGER
 412	Tail loss probe (TLP) converts RTOs occurring due to tail
 413	losses into fast recovery (draft-ietf-tcpm-rack). Note that
 414	TLP requires RACK to function properly (see tcp_recovery below)
 415
 416	Possible values:
 417
 418		- 0 disables TLP
 419		- 3 or 4 enables TLP
 420
 421	Default: 3
 422
 423tcp_ecn - INTEGER
 424	Control use of Explicit Congestion Notification (ECN) by TCP.
 425	ECN is used only when both ends of the TCP connection indicate
 426	support for it.  This feature is useful in avoiding losses due
 427	to congestion by allowing supporting routers to signal
 428	congestion before having to drop packets.
 429
 430	Possible values are:
 431
 432		=  =====================================================
 433		0  Disable ECN.  Neither initiate nor accept ECN.
 434		1  Enable ECN when requested by incoming connections and
 435		   also request ECN on outgoing connection attempts.
 436		2  Enable ECN when requested by incoming connections
 437		   but do not request ECN on outgoing connections.
 438		=  =====================================================
 439
 440	Default: 2
 441
 442tcp_ecn_fallback - BOOLEAN
 443	If the kernel detects that ECN connection misbehaves, enable fall
 444	back to non-ECN. Currently, this knob implements the fallback
 445	from RFC3168, section 6.1.1.1., but we reserve that in future,
 446	additional detection mechanisms could be implemented under this
 447	knob. The value	is not used, if tcp_ecn or per route (or congestion
 448	control) ECN settings are disabled.
 449
 450	Default: 1 (fallback enabled)
 451
 452tcp_fack - BOOLEAN
 453	This is a legacy option, it has no effect anymore.
 454
 455tcp_fin_timeout - INTEGER
 456	The length of time an orphaned (no longer referenced by any
 457	application) connection will remain in the FIN_WAIT_2 state
 458	before it is aborted at the local end.  While a perfectly
 459	valid "receive only" state for an un-orphaned connection, an
 460	orphaned connection in FIN_WAIT_2 state could otherwise wait
 461	forever for the remote to close its end of the connection.
 462
 463	Cf. tcp_max_orphans
 464
 465	Default: 60 seconds
 466
 467tcp_frto - INTEGER
 468	Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
 469	F-RTO is an enhanced recovery algorithm for TCP retransmission
 470	timeouts.  It is particularly beneficial in networks where the
 471	RTT fluctuates (e.g., wireless). F-RTO is sender-side only
 472	modification. It does not require any support from the peer.
 473
 474	By default it's enabled with a non-zero value. 0 disables F-RTO.
 475
 476tcp_fwmark_accept - BOOLEAN
 477	If set, incoming connections to listening sockets that do not have a
 478	socket mark will set the mark of the accepting socket to the fwmark of
 479	the incoming SYN packet. This will cause all packets on that connection
 480	(starting from the first SYNACK) to be sent with that fwmark. The
 481	listening socket's mark is unchanged. Listening sockets that already
 482	have a fwmark set via setsockopt(SOL_SOCKET, SO_MARK, ...) are
 483	unaffected.
 484
 485	Default: 0
 486
 487tcp_invalid_ratelimit - INTEGER
 488	Limit the maximal rate for sending duplicate acknowledgments
 489	in response to incoming TCP packets that are for an existing
 490	connection but that are invalid due to any of these reasons:
 491
 492	  (a) out-of-window sequence number,
 493	  (b) out-of-window acknowledgment number, or
 494	  (c) PAWS (Protection Against Wrapped Sequence numbers) check failure
 495
 496	This can help mitigate simple "ack loop" DoS attacks, wherein
 497	a buggy or malicious middlebox or man-in-the-middle can
 498	rewrite TCP header fields in manner that causes each endpoint
 499	to think that the other is sending invalid TCP segments, thus
 500	causing each side to send an unterminating stream of duplicate
 501	acknowledgments for invalid segments.
 502
 503	Using 0 disables rate-limiting of dupacks in response to
 504	invalid segments; otherwise this value specifies the minimal
 505	space between sending such dupacks, in milliseconds.
 506
 507	Default: 500 (milliseconds).
 508
 509tcp_keepalive_time - INTEGER
 510	How often TCP sends out keepalive messages when keepalive is enabled.
 511	Default: 2hours.
 512
 513tcp_keepalive_probes - INTEGER
 514	How many keepalive probes TCP sends out, until it decides that the
 515	connection is broken. Default value: 9.
 516
 517tcp_keepalive_intvl - INTEGER
 518	How frequently the probes are send out. Multiplied by
 519	tcp_keepalive_probes it is time to kill not responding connection,
 520	after probes started. Default value: 75sec i.e. connection
 521	will be aborted after ~11 minutes of retries.
 522
 523tcp_l3mdev_accept - BOOLEAN
 524	Enables child sockets to inherit the L3 master device index.
 525	Enabling this option allows a "global" listen socket to work
 526	across L3 master domains (e.g., VRFs) with connected sockets
 527	derived from the listen socket to be bound to the L3 domain in
 528	which the packets originated. Only valid when the kernel was
 529	compiled with CONFIG_NET_L3_MASTER_DEV.
 530
 531	Default: 0 (disabled)
 532
 533tcp_low_latency - BOOLEAN
 534	This is a legacy option, it has no effect anymore.
 535
 536tcp_max_orphans - INTEGER
 537	Maximal number of TCP sockets not attached to any user file handle,
 538	held by system.	If this number is exceeded orphaned connections are
 539	reset immediately and warning is printed. This limit exists
 540	only to prevent simple DoS attacks, you _must_ not rely on this
 541	or lower the limit artificially, but rather increase it
 542	(probably, after increasing installed memory),
 543	if network conditions require more than default value,
 544	and tune network services to linger and kill such states
 545	more aggressively. Let me to remind again: each orphan eats
 546	up to ~64K of unswappable memory.
 547
 548tcp_max_syn_backlog - INTEGER
 549	Maximal number of remembered connection requests (SYN_RECV),
 550	which have not received an acknowledgment from connecting client.
 551
 552	This is a per-listener limit.
 553
 554	The minimal value is 128 for low memory machines, and it will
 555	increase in proportion to the memory of machine.
 556
 557	If server suffers from overload, try increasing this number.
 558
 559	Remember to also check /proc/sys/net/core/somaxconn
 560	A SYN_RECV request socket consumes about 304 bytes of memory.
 561
 562tcp_max_tw_buckets - INTEGER
 563	Maximal number of timewait sockets held by system simultaneously.
 564	If this number is exceeded time-wait socket is immediately destroyed
 565	and warning is printed. This limit exists only to prevent
 566	simple DoS attacks, you _must_ not lower the limit artificially,
 567	but rather increase it (probably, after increasing installed memory),
 568	if network conditions require more than default value.
 569
 570tcp_mem - vector of 3 INTEGERs: min, pressure, max
 571	min: below this number of pages TCP is not bothered about its
 572	memory appetite.
 573
 574	pressure: when amount of memory allocated by TCP exceeds this number
 575	of pages, TCP moderates its memory consumption and enters memory
 576	pressure mode, which is exited when memory consumption falls
 577	under "min".
 578
 579	max: number of pages allowed for queueing by all TCP sockets.
 580
 581	Defaults are calculated at boot time from amount of available
 582	memory.
 583
 584tcp_min_rtt_wlen - INTEGER
 585	The window length of the windowed min filter to track the minimum RTT.
 586	A shorter window lets a flow more quickly pick up new (higher)
 587	minimum RTT when it is moved to a longer path (e.g., due to traffic
 588	engineering). A longer window makes the filter more resistant to RTT
 589	inflations such as transient congestion. The unit is seconds.
 590
 591	Possible values: 0 - 86400 (1 day)
 592
 593	Default: 300
 594
 595tcp_moderate_rcvbuf - BOOLEAN
 596	If set, TCP performs receive buffer auto-tuning, attempting to
 597	automatically size the buffer (no greater than tcp_rmem[2]) to
 598	match the size required by the path for full throughput.  Enabled by
 599	default.
 600
 601tcp_mtu_probing - INTEGER
 602	Controls TCP Packetization-Layer Path MTU Discovery.  Takes three
 603	values:
 604
 605	- 0 - Disabled
 606	- 1 - Disabled by default, enabled when an ICMP black hole detected
 607	- 2 - Always enabled, use initial MSS of tcp_base_mss.
 608
 609tcp_probe_interval - UNSIGNED INTEGER
 610	Controls how often to start TCP Packetization-Layer Path MTU
 611	Discovery reprobe. The default is reprobing every 10 minutes as
 612	per RFC4821.
 613
 614tcp_probe_threshold - INTEGER
 615	Controls when TCP Packetization-Layer Path MTU Discovery probing
 616	will stop in respect to the width of search range in bytes. Default
 617	is 8 bytes.
 618
 619tcp_no_metrics_save - BOOLEAN
 620	By default, TCP saves various connection metrics in the route cache
 621	when the connection closes, so that connections established in the
 622	near future can use these to set initial conditions.  Usually, this
 623	increases overall performance, but may sometimes cause performance
 624	degradation.  If set, TCP will not cache metrics on closing
 625	connections.
 626
 627tcp_no_ssthresh_metrics_save - BOOLEAN
 628	Controls whether TCP saves ssthresh metrics in the route cache.
 629
 630	Default is 1, which disables ssthresh metrics.
 631
 632tcp_orphan_retries - INTEGER
 633	This value influences the timeout of a locally closed TCP connection,
 634	when RTO retransmissions remain unacknowledged.
 635	See tcp_retries2 for more details.
 636
 637	The default value is 8.
 638
 639	If your machine is a loaded WEB server,
 640	you should think about lowering this value, such sockets
 641	may consume significant resources. Cf. tcp_max_orphans.
 642
 643tcp_recovery - INTEGER
 644	This value is a bitmap to enable various experimental loss recovery
 645	features.
 646
 647	=========   =============================================================
 648	RACK: 0x1   enables the RACK loss detection for fast detection of lost
 649		    retransmissions and tail drops. It also subsumes and disables
 650		    RFC6675 recovery for SACK connections.
 651
 652	RACK: 0x2   makes RACK's reordering window static (min_rtt/4).
 653
 654	RACK: 0x4   disables RACK's DUPACK threshold heuristic
 655	=========   =============================================================
 656
 657	Default: 0x1
 658
 659tcp_reflect_tos - BOOLEAN
 660	For listening sockets, reuse the DSCP value of the initial SYN message
 661	for outgoing packets. This allows to have both directions of a TCP
 662	stream to use the same DSCP value, assuming DSCP remains unchanged for
 663	the lifetime of the connection.
 664
 665	This options affects both IPv4 and IPv6.
 666
 667	Default: 0 (disabled)
 668
 669tcp_reordering - INTEGER
 670	Initial reordering level of packets in a TCP stream.
 671	TCP stack can then dynamically adjust flow reordering level
 672	between this initial value and tcp_max_reordering
 673
 674	Default: 3
 675
 676tcp_max_reordering - INTEGER
 677	Maximal reordering level of packets in a TCP stream.
 678	300 is a fairly conservative value, but you might increase it
 679	if paths are using per packet load balancing (like bonding rr mode)
 680
 681	Default: 300
 682
 683tcp_retrans_collapse - BOOLEAN
 684	Bug-to-bug compatibility with some broken printers.
 685	On retransmit try to send bigger packets to work around bugs in
 686	certain TCP stacks.
 687
 688tcp_retries1 - INTEGER
 689	This value influences the time, after which TCP decides, that
 690	something is wrong due to unacknowledged RTO retransmissions,
 691	and reports this suspicion to the network layer.
 692	See tcp_retries2 for more details.
 693
 694	RFC 1122 recommends at least 3 retransmissions, which is the
 695	default.
 696
 697tcp_retries2 - INTEGER
 698	This value influences the timeout of an alive TCP connection,
 699	when RTO retransmissions remain unacknowledged.
 700	Given a value of N, a hypothetical TCP connection following
 701	exponential backoff with an initial RTO of TCP_RTO_MIN would
 702	retransmit N times before killing the connection at the (N+1)th RTO.
 703
 704	The default value of 15 yields a hypothetical timeout of 924.6
 705	seconds and is a lower bound for the effective timeout.
 706	TCP will effectively time out at the first RTO which exceeds the
 707	hypothetical timeout.
 708
 709	RFC 1122 recommends at least 100 seconds for the timeout,
 710	which corresponds to a value of at least 8.
 711
 712tcp_rfc1337 - BOOLEAN
 713	If set, the TCP stack behaves conforming to RFC1337. If unset,
 714	we are not conforming to RFC, but prevent TCP TIME_WAIT
 715	assassination.
 716
 717	Default: 0
 718
 719tcp_rmem - vector of 3 INTEGERs: min, default, max
 720	min: Minimal size of receive buffer used by TCP sockets.
 721	It is guaranteed to each TCP socket, even under moderate memory
 722	pressure.
 723
 724	Default: 4K
 725
 726	default: initial size of receive buffer used by TCP sockets.
 727	This value overrides net.core.rmem_default used by other protocols.
 728	Default: 131072 bytes.
 729	This value results in initial window of 65535.
 730
 731	max: maximal size of receive buffer allowed for automatically
 732	selected receiver buffers for TCP socket. This value does not override
 733	net.core.rmem_max.  Calling setsockopt() with SO_RCVBUF disables
 734	automatic tuning of that socket's receive buffer size, in which
 735	case this value is ignored.
 736	Default: between 131072 and 6MB, depending on RAM size.
 737
 738tcp_sack - BOOLEAN
 739	Enable select acknowledgments (SACKS).
 740
 741tcp_comp_sack_delay_ns - LONG INTEGER
 742	TCP tries to reduce number of SACK sent, using a timer
 743	based on 5% of SRTT, capped by this sysctl, in nano seconds.
 744	The default is 1ms, based on TSO autosizing period.
 745
 746	Default : 1,000,000 ns (1 ms)
 747
 748tcp_comp_sack_slack_ns - LONG INTEGER
 749	This sysctl control the slack used when arming the
 750	timer used by SACK compression. This gives extra time
 751	for small RTT flows, and reduces system overhead by allowing
 752	opportunistic reduction of timer interrupts.
 753
 754	Default : 100,000 ns (100 us)
 755
 756tcp_comp_sack_nr - INTEGER
 757	Max number of SACK that can be compressed.
 758	Using 0 disables SACK compression.
 759
 760	Default : 44
 761
 762tcp_backlog_ack_defer - BOOLEAN
 763	If set, user thread processing socket backlog tries sending
 764	one ACK for the whole queue. This helps to avoid potential
 765	long latencies at end of a TCP socket syscall.
 766
 767	Default : true
 768
 769tcp_slow_start_after_idle - BOOLEAN
 770	If set, provide RFC2861 behavior and time out the congestion
 771	window after an idle period.  An idle period is defined at
 772	the current RTO.  If unset, the congestion window will not
 773	be timed out after an idle period.
 774
 775	Default: 1
 776
 777tcp_stdurg - BOOLEAN
 778	Use the Host requirements interpretation of the TCP urgent pointer field.
 779	Most hosts use the older BSD interpretation, so if you turn this on
 780	Linux might not communicate correctly with them.
 781
 782	Default: FALSE
 783
 784tcp_synack_retries - INTEGER
 785	Number of times SYNACKs for a passive TCP connection attempt will
 786	be retransmitted. Should not be higher than 255. Default value
 787	is 5, which corresponds to 31seconds till the last retransmission
 788	with the current initial RTO of 1second. With this the final timeout
 789	for a passive TCP connection will happen after 63seconds.
 790
 791tcp_syncookies - INTEGER
 792	Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
 793	Send out syncookies when the syn backlog queue of a socket
 794	overflows. This is to prevent against the common 'SYN flood attack'
 795	Default: 1
 796
 797	Note, that syncookies is fallback facility.
 798	It MUST NOT be used to help highly loaded servers to stand
 799	against legal connection rate. If you see SYN flood warnings
 800	in your logs, but investigation	shows that they occur
 801	because of overload with legal connections, you should tune
 802	another parameters until this warning disappear.
 803	See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
 804
 805	syncookies seriously violate TCP protocol, do not allow
 806	to use TCP extensions, can result in serious degradation
 807	of some services (f.e. SMTP relaying), visible not by you,
 808	but your clients and relays, contacting you. While you see
 809	SYN flood warnings in logs not being really flooded, your server
 810	is seriously misconfigured.
 811
 812	If you want to test which effects syncookies have to your
 813	network connections you can set this knob to 2 to enable
 814	unconditionally generation of syncookies.
 815
 816tcp_migrate_req - BOOLEAN
 817	The incoming connection is tied to a specific listening socket when
 818	the initial SYN packet is received during the three-way handshake.
 819	When a listener is closed, in-flight request sockets during the
 820	handshake and established sockets in the accept queue are aborted.
 821
 822	If the listener has SO_REUSEPORT enabled, other listeners on the
 823	same port should have been able to accept such connections. This
 824	option makes it possible to migrate such child sockets to another
 825	listener after close() or shutdown().
 826
 827	The BPF_SK_REUSEPORT_SELECT_OR_MIGRATE type of eBPF program should
 828	usually be used to define the policy to pick an alive listener.
 829	Otherwise, the kernel will randomly pick an alive listener only if
 830	this option is enabled.
 831
 832	Note that migration between listeners with different settings may
 833	crash applications. Let's say migration happens from listener A to
 834	B, and only B has TCP_SAVE_SYN enabled. B cannot read SYN data from
 835	the requests migrated from A. To avoid such a situation, cancel
 836	migration by returning SK_DROP in the type of eBPF program, or
 837	disable this option.
 838
 839	Default: 0
 840
 841tcp_fastopen - INTEGER
 842	Enable TCP Fast Open (RFC7413) to send and accept data in the opening
 843	SYN packet.
 844
 845	The client support is enabled by flag 0x1 (on by default). The client
 846	then must use sendmsg() or sendto() with the MSG_FASTOPEN flag,
 847	rather than connect() to send data in SYN.
 848
 849	The server support is enabled by flag 0x2 (off by default). Then
 850	either enable for all listeners with another flag (0x400) or
 851	enable individual listeners via TCP_FASTOPEN socket option with
 852	the option value being the length of the syn-data backlog.
 853
 854	The values (bitmap) are
 855
 856	=====  ======== ======================================================
 857	  0x1  (client) enables sending data in the opening SYN on the client.
 858	  0x2  (server) enables the server support, i.e., allowing data in
 859			a SYN packet to be accepted and passed to the
 860			application before 3-way handshake finishes.
 861	  0x4  (client) send data in the opening SYN regardless of cookie
 862			availability and without a cookie option.
 863	0x200  (server) accept data-in-SYN w/o any cookie option present.
 864	0x400  (server) enable all listeners to support Fast Open by
 865			default without explicit TCP_FASTOPEN socket option.
 866	=====  ======== ======================================================
 867
 868	Default: 0x1
 869
 870	Note that additional client or server features are only
 871	effective if the basic support (0x1 and 0x2) are enabled respectively.
 872
 873tcp_fastopen_blackhole_timeout_sec - INTEGER
 874	Initial time period in second to disable Fastopen on active TCP sockets
 875	when a TFO firewall blackhole issue happens.
 876	This time period will grow exponentially when more blackhole issues
 877	get detected right after Fastopen is re-enabled and will reset to
 878	initial value when the blackhole issue goes away.
 879	0 to disable the blackhole detection.
 880
 881	By default, it is set to 0 (feature is disabled).
 882
 883tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
 884	The list consists of a primary key and an optional backup key. The
 885	primary key is used for both creating and validating cookies, while the
 886	optional backup key is only used for validating cookies. The purpose of
 887	the backup key is to maximize TFO validation when keys are rotated.
 888
 889	A randomly chosen primary key may be configured by the kernel if
 890	the tcp_fastopen sysctl is set to 0x400 (see above), or if the
 891	TCP_FASTOPEN setsockopt() optname is set and a key has not been
 892	previously configured via sysctl. If keys are configured via
 893	setsockopt() by using the TCP_FASTOPEN_KEY optname, then those
 894	per-socket keys will be used instead of any keys that are specified via
 895	sysctl.
 896
 897	A key is specified as 4 8-digit hexadecimal integers which are separated
 898	by a '-' as: xxxxxxxx-xxxxxxxx-xxxxxxxx-xxxxxxxx. Leading zeros may be
 899	omitted. A primary and a backup key may be specified by separating them
 900	by a comma. If only one key is specified, it becomes the primary key and
 901	any previously configured backup keys are removed.
 902
 903tcp_syn_retries - INTEGER
 904	Number of times initial SYNs for an active TCP connection attempt
 905	will be retransmitted. Should not be higher than 127. Default value
 906	is 6, which corresponds to 67seconds (with tcp_syn_linear_timeouts = 4)
 907	till the last retransmission with the current initial RTO of 1second.
 908	With this the final timeout for an active TCP connection attempt
 909	will happen after 131seconds.
 910
 911tcp_timestamps - INTEGER
 912	Enable timestamps as defined in RFC1323.
 913
 914	- 0: Disabled.
 915	- 1: Enable timestamps as defined in RFC1323 and use random offset for
 916	  each connection rather than only using the current time.
 917	- 2: Like 1, but without random offsets.
 918
 919	Default: 1
 920
 921tcp_min_tso_segs - INTEGER
 922	Minimal number of segments per TSO frame.
 923
 924	Since linux-3.12, TCP does an automatic sizing of TSO frames,
 925	depending on flow rate, instead of filling 64Kbytes packets.
 926	For specific usages, it's possible to force TCP to build big
 927	TSO frames. Note that TCP stack might split too big TSO packets
 928	if available window is too small.
 929
 930	Default: 2
 931
 932tcp_tso_rtt_log - INTEGER
 933	Adjustment of TSO packet sizes based on min_rtt
 934
 935	Starting from linux-5.18, TCP autosizing can be tweaked
 936	for flows having small RTT.
 937
 938	Old autosizing was splitting the pacing budget to send 1024 TSO
 939	per second.
 940
 941	tso_packet_size = sk->sk_pacing_rate / 1024;
 942
 943	With the new mechanism, we increase this TSO sizing using:
 944
 945	distance = min_rtt_usec / (2^tcp_tso_rtt_log)
 946	tso_packet_size += gso_max_size >> distance;
 947
 948	This means that flows between very close hosts can use bigger
 949	TSO packets, reducing their cpu costs.
 950
 951	If you want to use the old autosizing, set this sysctl to 0.
 952
 953	Default: 9  (2^9 = 512 usec)
 954
 955tcp_pacing_ss_ratio - INTEGER
 956	sk->sk_pacing_rate is set by TCP stack using a ratio applied
 957	to current rate. (current_rate = cwnd * mss / srtt)
 958	If TCP is in slow start, tcp_pacing_ss_ratio is applied
 959	to let TCP probe for bigger speeds, assuming cwnd can be
 960	doubled every other RTT.
 961
 962	Default: 200
 963
 964tcp_pacing_ca_ratio - INTEGER
 965	sk->sk_pacing_rate is set by TCP stack using a ratio applied
 966	to current rate. (current_rate = cwnd * mss / srtt)
 967	If TCP is in congestion avoidance phase, tcp_pacing_ca_ratio
 968	is applied to conservatively probe for bigger throughput.
 969
 970	Default: 120
 971
 972tcp_syn_linear_timeouts - INTEGER
 973	The number of times for an active TCP connection to retransmit SYNs with
 974	a linear backoff timeout before defaulting to an exponential backoff
 975	timeout. This has no effect on SYNACK at the passive TCP side.
 976
 977	With an initial RTO of 1 and tcp_syn_linear_timeouts = 4 we would
 978	expect SYN RTOs to be: 1, 1, 1, 1, 1, 2, 4, ... (4 linear timeouts,
 979	and the first exponential backoff using 2^0 * initial_RTO).
 980	Default: 4
 981
 982tcp_tso_win_divisor - INTEGER
 983	This allows control over what percentage of the congestion window
 984	can be consumed by a single TSO frame.
 985	The setting of this parameter is a choice between burstiness and
 986	building larger TSO frames.
 987
 988	Default: 3
 989
 990tcp_tw_reuse - INTEGER
 991	Enable reuse of TIME-WAIT sockets for new connections when it is
 992	safe from protocol viewpoint.
 993
 994	- 0 - disable
 995	- 1 - global enable
 996	- 2 - enable for loopback traffic only
 997
 998	It should not be changed without advice/request of technical
 999	experts.
1000
1001	Default: 2
1002
1003tcp_window_scaling - BOOLEAN
1004	Enable window scaling as defined in RFC1323.
1005
1006tcp_shrink_window - BOOLEAN
1007	This changes how the TCP receive window is calculated.
1008
1009	RFC 7323, section 2.4, says there are instances when a retracted
1010	window can be offered, and that TCP implementations MUST ensure
1011	that they handle a shrinking window, as specified in RFC 1122.
1012
1013	- 0 - Disabled.	The window is never shrunk.
1014	- 1 - Enabled.	The window is shrunk when necessary to remain within
1015			the memory limit set by autotuning (sk_rcvbuf).
1016			This only occurs if a non-zero receive window
1017			scaling factor is also in effect.
1018
1019	Default: 0
1020
1021tcp_wmem - vector of 3 INTEGERs: min, default, max
1022	min: Amount of memory reserved for send buffers for TCP sockets.
1023	Each TCP socket has rights to use it due to fact of its birth.
1024
1025	Default: 4K
1026
1027	default: initial size of send buffer used by TCP sockets.  This
1028	value overrides net.core.wmem_default used by other protocols.
1029
1030	It is usually lower than net.core.wmem_default.
1031
1032	Default: 16K
1033
1034	max: Maximal amount of memory allowed for automatically tuned
1035	send buffers for TCP sockets. This value does not override
1036	net.core.wmem_max.  Calling setsockopt() with SO_SNDBUF disables
1037	automatic tuning of that socket's send buffer size, in which case
1038	this value is ignored.
1039
1040	Default: between 64K and 4MB, depending on RAM size.
1041
1042tcp_notsent_lowat - UNSIGNED INTEGER
1043	A TCP socket can control the amount of unsent bytes in its write queue,
1044	thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
1045	reports POLLOUT events if the amount of unsent bytes is below a per
1046	socket value, and if the write queue is not full. sendmsg() will
1047	also not add new buffers if the limit is hit.
1048
1049	This global variable controls the amount of unsent data for
1050	sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
1051	to the global variable has immediate effect.
1052
1053	Default: UINT_MAX (0xFFFFFFFF)
1054
1055tcp_workaround_signed_windows - BOOLEAN
1056	If set, assume no receipt of a window scaling option means the
1057	remote TCP is broken and treats the window as a signed quantity.
1058	If unset, assume the remote TCP is not broken even if we do
1059	not receive a window scaling option from them.
1060
1061	Default: 0
1062
1063tcp_thin_linear_timeouts - BOOLEAN
1064	Enable dynamic triggering of linear timeouts for thin streams.
1065	If set, a check is performed upon retransmission by timeout to
1066	determine if the stream is thin (less than 4 packets in flight).
1067	As long as the stream is found to be thin, up to 6 linear
1068	timeouts may be performed before exponential backoff mode is
1069	initiated. This improves retransmission latency for
1070	non-aggressive thin streams, often found to be time-dependent.
1071	For more information on thin streams, see
1072	Documentation/networking/tcp-thin.rst
1073
1074	Default: 0
1075
1076tcp_limit_output_bytes - INTEGER
1077	Controls TCP Small Queue limit per tcp socket.
1078	TCP bulk sender tends to increase packets in flight until it
1079	gets losses notifications. With SNDBUF autotuning, this can
1080	result in a large amount of packets queued on the local machine
1081	(e.g.: qdiscs, CPU backlog, or device) hurting latency of other
1082	flows, for typical pfifo_fast qdiscs.  tcp_limit_output_bytes
1083	limits the number of bytes on qdisc or device to reduce artificial
1084	RTT/cwnd and reduce bufferbloat.
1085
1086	Default: 1048576 (16 * 65536)
1087
1088tcp_challenge_ack_limit - INTEGER
1089	Limits number of Challenge ACK sent per second, as recommended
1090	in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
1091	Note that this per netns rate limit can allow some side channel
1092	attacks and probably should not be enabled.
1093	TCP stack implements per TCP socket limits anyway.
1094	Default: INT_MAX (unlimited)
1095
1096tcp_ehash_entries - INTEGER
1097	Show the number of hash buckets for TCP sockets in the current
1098	networking namespace.
1099
1100	A negative value means the networking namespace does not own its
1101	hash buckets and shares the initial networking namespace's one.
1102
1103tcp_child_ehash_entries - INTEGER
1104	Control the number of hash buckets for TCP sockets in the child
1105	networking namespace, which must be set before clone() or unshare().
1106
1107	If the value is not 0, the kernel uses a value rounded up to 2^n
1108	as the actual hash bucket size.  0 is a special value, meaning
1109	the child networking namespace will share the initial networking
1110	namespace's hash buckets.
1111
1112	Note that the child will use the global one in case the kernel
1113	fails to allocate enough memory.  In addition, the global hash
1114	buckets are spread over available NUMA nodes, but the allocation
1115	of the child hash table depends on the current process's NUMA
1116	policy, which could result in performance differences.
1117
1118	Note also that the default value of tcp_max_tw_buckets and
1119	tcp_max_syn_backlog depend on the hash bucket size.
1120
1121	Possible values: 0, 2^n (n: 0 - 24 (16Mi))
1122
1123	Default: 0
1124
1125tcp_plb_enabled - BOOLEAN
1126	If set and the underlying congestion control (e.g. DCTCP) supports
1127	and enables PLB feature, TCP PLB (Protective Load Balancing) is
1128	enabled. PLB is described in the following paper:
1129	https://doi.org/10.1145/3544216.3544226. Based on PLB parameters,
1130	upon sensing sustained congestion, TCP triggers a change in
1131	flow label field for outgoing IPv6 packets. A change in flow label
1132	field potentially changes the path of outgoing packets for switches
1133	that use ECMP/WCMP for routing.
1134
1135	PLB changes socket txhash which results in a change in IPv6 Flow Label
1136	field, and currently no-op for IPv4 headers. It is possible
1137	to apply PLB for IPv4 with other network header fields (e.g. TCP
1138	or IPv4 options) or using encapsulation where outer header is used
1139	by switches to determine next hop. In either case, further host
1140	and switch side changes will be needed.
1141
1142	When set, PLB assumes that congestion signal (e.g. ECN) is made
1143	available and used by congestion control module to estimate a
1144	congestion measure (e.g. ce_ratio). PLB needs a congestion measure to
1145	make repathing decisions.
1146
1147	Default: FALSE
1148
1149tcp_plb_idle_rehash_rounds - INTEGER
1150	Number of consecutive congested rounds (RTT) seen after which
1151	a rehash can be performed, given there are no packets in flight.
1152	This is referred to as M in PLB paper:
1153	https://doi.org/10.1145/3544216.3544226.
1154
1155	Possible Values: 0 - 31
1156
1157	Default: 3
1158
1159tcp_plb_rehash_rounds - INTEGER
1160	Number of consecutive congested rounds (RTT) seen after which
1161	a forced rehash can be performed. Be careful when setting this
1162	parameter, as a small value increases the risk of retransmissions.
1163	This is referred to as N in PLB paper:
1164	https://doi.org/10.1145/3544216.3544226.
1165
1166	Possible Values: 0 - 31
1167
1168	Default: 12
1169
1170tcp_plb_suspend_rto_sec - INTEGER
1171	Time, in seconds, to suspend PLB in event of an RTO. In order to avoid
1172	having PLB repath onto a connectivity "black hole", after an RTO a TCP
1173	connection suspends PLB repathing for a random duration between 1x and
1174	2x of this parameter. Randomness is added to avoid concurrent rehashing
1175	of multiple TCP connections. This should be set corresponding to the
1176	amount of time it takes to repair a failed link.
1177
1178	Possible Values: 0 - 255
1179
1180	Default: 60
1181
1182tcp_plb_cong_thresh - INTEGER
1183	Fraction of packets marked with congestion over a round (RTT) to
1184	tag that round as congested. This is referred to as K in the PLB paper:
1185	https://doi.org/10.1145/3544216.3544226.
1186
1187	The 0-1 fraction range is mapped to 0-256 range to avoid floating
1188	point operations. For example, 128 means that if at least 50% of
1189	the packets in a round were marked as congested then the round
1190	will be tagged as congested.
1191
1192	Setting threshold to 0 means that PLB repaths every RTT regardless
1193	of congestion. This is not intended behavior for PLB and should be
1194	used only for experimentation purpose.
1195
1196	Possible Values: 0 - 256
1197
1198	Default: 128
1199
1200tcp_pingpong_thresh - INTEGER
1201	The number of estimated data replies sent for estimated incoming data
1202	requests that must happen before TCP considers that a connection is a
1203	"ping-pong" (request-response) connection for which delayed
1204	acknowledgments can provide benefits.
1205
1206	This threshold is 1 by default, but some applications may need a higher
1207	threshold for optimal performance.
1208
1209	Possible Values: 1 - 255
1210
1211	Default: 1
1212
1213tcp_rto_min_us - INTEGER
1214	Minimal TCP retransmission timeout (in microseconds). Note that the
1215	rto_min route option has the highest precedence for configuring this
1216	setting, followed by the TCP_BPF_RTO_MIN socket option, followed by
1217	this tcp_rto_min_us sysctl.
1218
1219	The recommended practice is to use a value less or equal to 200000
1220	microseconds.
1221
1222	Possible Values: 1 - INT_MAX
1223
1224	Default: 200000
1225
1226UDP variables
1227=============
1228
1229udp_l3mdev_accept - BOOLEAN
1230	Enabling this option allows a "global" bound socket to work
1231	across L3 master domains (e.g., VRFs) with packets capable of
1232	being received regardless of the L3 domain in which they
1233	originated. Only valid when the kernel was compiled with
1234	CONFIG_NET_L3_MASTER_DEV.
1235
1236	Default: 0 (disabled)
1237
1238udp_mem - vector of 3 INTEGERs: min, pressure, max
1239	Number of pages allowed for queueing by all UDP sockets.
1240
1241	min: Number of pages allowed for queueing by all UDP sockets.
1242
1243	pressure: This value was introduced to follow format of tcp_mem.
1244
1245	max: This value was introduced to follow format of tcp_mem.
1246
1247	Default is calculated at boot time from amount of available memory.
1248
1249udp_rmem_min - INTEGER
1250	Minimal size of receive buffer used by UDP sockets in moderation.
1251	Each UDP socket is able to use the size for receiving data, even if
1252	total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
1253
1254	Default: 4K
1255
1256udp_wmem_min - INTEGER
1257	UDP does not have tx memory accounting and this tunable has no effect.
1258
1259udp_hash_entries - INTEGER
1260	Show the number of hash buckets for UDP sockets in the current
1261	networking namespace.
1262
1263	A negative value means the networking namespace does not own its
1264	hash buckets and shares the initial networking namespace's one.
1265
1266udp_child_ehash_entries - INTEGER
1267	Control the number of hash buckets for UDP sockets in the child
1268	networking namespace, which must be set before clone() or unshare().
1269
1270	If the value is not 0, the kernel uses a value rounded up to 2^n
1271	as the actual hash bucket size.  0 is a special value, meaning
1272	the child networking namespace will share the initial networking
1273	namespace's hash buckets.
1274
1275	Note that the child will use the global one in case the kernel
1276	fails to allocate enough memory.  In addition, the global hash
1277	buckets are spread over available NUMA nodes, but the allocation
1278	of the child hash table depends on the current process's NUMA
1279	policy, which could result in performance differences.
1280
1281	Possible values: 0, 2^n (n: 7 (128) - 16 (64K))
1282
1283	Default: 0
1284
1285
1286RAW variables
1287=============
1288
1289raw_l3mdev_accept - BOOLEAN
1290	Enabling this option allows a "global" bound socket to work
1291	across L3 master domains (e.g., VRFs) with packets capable of
1292	being received regardless of the L3 domain in which they
1293	originated. Only valid when the kernel was compiled with
1294	CONFIG_NET_L3_MASTER_DEV.
1295
1296	Default: 1 (enabled)
1297
1298CIPSOv4 Variables
1299=================
1300
1301cipso_cache_enable - BOOLEAN
1302	If set, enable additions to and lookups from the CIPSO label mapping
1303	cache.  If unset, additions are ignored and lookups always result in a
1304	miss.  However, regardless of the setting the cache is still
1305	invalidated when required when means you can safely toggle this on and
1306	off and the cache will always be "safe".
1307
1308	Default: 1
1309
1310cipso_cache_bucket_size - INTEGER
1311	The CIPSO label cache consists of a fixed size hash table with each
1312	hash bucket containing a number of cache entries.  This variable limits
1313	the number of entries in each hash bucket; the larger the value is, the
1314	more CIPSO label mappings that can be cached.  When the number of
1315	entries in a given hash bucket reaches this limit adding new entries
1316	causes the oldest entry in the bucket to be removed to make room.
1317
1318	Default: 10
1319
1320cipso_rbm_optfmt - BOOLEAN
1321	Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
1322	the CIPSO draft specification (see Documentation/netlabel for details).
1323	This means that when set the CIPSO tag will be padded with empty
1324	categories in order to make the packet data 32-bit aligned.
1325
1326	Default: 0
1327
1328cipso_rbm_structvalid - BOOLEAN
1329	If set, do a very strict check of the CIPSO option when
1330	ip_options_compile() is called.  If unset, relax the checks done during
1331	ip_options_compile().  Either way is "safe" as errors are caught else
1332	where in the CIPSO processing code but setting this to 0 (False) should
1333	result in less work (i.e. it should be faster) but could cause problems
1334	with other implementations that require strict checking.
1335
1336	Default: 0
1337
1338IP Variables
1339============
1340
1341ip_local_port_range - 2 INTEGERS
1342	Defines the local port range that is used by TCP and UDP to
1343	choose the local port. The first number is the first, the
1344	second the last local port number.
1345	If possible, it is better these numbers have different parity
1346	(one even and one odd value).
1347	Must be greater than or equal to ip_unprivileged_port_start.
1348	The default values are 32768 and 60999 respectively.
1349
1350ip_local_reserved_ports - list of comma separated ranges
1351	Specify the ports which are reserved for known third-party
1352	applications. These ports will not be used by automatic port
1353	assignments (e.g. when calling connect() or bind() with port
1354	number 0). Explicit port allocation behavior is unchanged.
1355
1356	The format used for both input and output is a comma separated
1357	list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
1358	10). Writing to the file will clear all previously reserved
1359	ports and update the current list with the one given in the
1360	input.
1361
1362	Note that ip_local_port_range and ip_local_reserved_ports
1363	settings are independent and both are considered by the kernel
1364	when determining which ports are available for automatic port
1365	assignments.
1366
1367	You can reserve ports which are not in the current
1368	ip_local_port_range, e.g.::
1369
1370	    $ cat /proc/sys/net/ipv4/ip_local_port_range
1371	    32000	60999
1372	    $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
1373	    8080,9148
1374
1375	although this is redundant. However such a setting is useful
1376	if later the port range is changed to a value that will
1377	include the reserved ports. Also keep in mind, that overlapping
1378	of these ranges may affect probability of selecting ephemeral
1379	ports which are right after block of reserved ports.
1380
1381	Default: Empty
1382
1383ip_unprivileged_port_start - INTEGER
1384	This is a per-namespace sysctl.  It defines the first
1385	unprivileged port in the network namespace.  Privileged ports
1386	require root or CAP_NET_BIND_SERVICE in order to bind to them.
1387	To disable all privileged ports, set this to 0.  They must not
1388	overlap with the ip_local_port_range.
1389
1390	Default: 1024
1391
1392ip_nonlocal_bind - BOOLEAN
1393	If set, allows processes to bind() to non-local IP addresses,
1394	which can be quite useful - but may break some applications.
1395
1396	Default: 0
1397
1398ip_autobind_reuse - BOOLEAN
1399	By default, bind() does not select the ports automatically even if
1400	the new socket and all sockets bound to the port have SO_REUSEADDR.
1401	ip_autobind_reuse allows bind() to reuse the port and this is useful
1402	when you use bind()+connect(), but may break some applications.
1403	The preferred solution is to use IP_BIND_ADDRESS_NO_PORT and this
1404	option should only be set by experts.
1405	Default: 0
1406
1407ip_dynaddr - INTEGER
1408	If set non-zero, enables support for dynamic addresses.
1409	If set to a non-zero value larger than 1, a kernel log
1410	message will be printed when dynamic address rewriting
1411	occurs.
1412
1413	Default: 0
1414
1415ip_early_demux - BOOLEAN
1416	Optimize input packet processing down to one demux for
1417	certain kinds of local sockets.  Currently we only do this
1418	for established TCP and connected UDP sockets.
1419
1420	It may add an additional cost for pure routing workloads that
1421	reduces overall throughput, in such case you should disable it.
1422
1423	Default: 1
1424
1425ping_group_range - 2 INTEGERS
1426	Restrict ICMP_PROTO datagram sockets to users in the group range.
1427	The default is "1 0", meaning, that nobody (not even root) may
1428	create ping sockets.  Setting it to "100 100" would grant permissions
1429	to the single group. "0 4294967294" would enable it for the world, "100
1430	4294967294" would enable it for the users, but not daemons.
1431
1432tcp_early_demux - BOOLEAN
1433	Enable early demux for established TCP sockets.
1434
1435	Default: 1
1436
1437udp_early_demux - BOOLEAN
1438	Enable early demux for connected UDP sockets. Disable this if
1439	your system could experience more unconnected load.
1440
1441	Default: 1
1442
1443icmp_echo_ignore_all - BOOLEAN
1444	If set non-zero, then the kernel will ignore all ICMP ECHO
1445	requests sent to it.
1446
1447	Default: 0
1448
1449icmp_echo_enable_probe - BOOLEAN
1450        If set to one, then the kernel will respond to RFC 8335 PROBE
1451        requests sent to it.
1452
1453        Default: 0
1454
1455icmp_echo_ignore_broadcasts - BOOLEAN
1456	If set non-zero, then the kernel will ignore all ICMP ECHO and
1457	TIMESTAMP requests sent to it via broadcast/multicast.
1458
1459	Default: 1
1460
1461icmp_ratelimit - INTEGER
1462	Limit the maximal rates for sending ICMP packets whose type matches
1463	icmp_ratemask (see below) to specific targets.
1464	0 to disable any limiting,
1465	otherwise the minimal space between responses in milliseconds.
1466	Note that another sysctl, icmp_msgs_per_sec limits the number
1467	of ICMP packets	sent on all targets.
1468
1469	Default: 1000
1470
1471icmp_msgs_per_sec - INTEGER
1472	Limit maximal number of ICMP packets sent per second from this host.
1473	Only messages whose type matches icmp_ratemask (see below) are
1474	controlled by this limit. For security reasons, the precise count
1475	of messages per second is randomized.
1476
1477	Default: 1000
1478
1479icmp_msgs_burst - INTEGER
1480	icmp_msgs_per_sec controls number of ICMP packets sent per second,
1481	while icmp_msgs_burst controls the burst size of these packets.
1482	For security reasons, the precise burst size is randomized.
1483
1484	Default: 50
1485
1486icmp_ratemask - INTEGER
1487	Mask made of ICMP types for which rates are being limited.
1488
1489	Significant bits: IHGFEDCBA9876543210
1490
1491	Default mask:     0000001100000011000 (6168)
1492
1493	Bit definitions (see include/linux/icmp.h):
1494
1495		= =========================
1496		0 Echo Reply
1497		3 Destination Unreachable [1]_
1498		4 Source Quench [1]_
1499		5 Redirect
1500		8 Echo Request
1501		B Time Exceeded [1]_
1502		C Parameter Problem [1]_
1503		D Timestamp Request
1504		E Timestamp Reply
1505		F Info Request
1506		G Info Reply
1507		H Address Mask Request
1508		I Address Mask Reply
1509		= =========================
1510
1511	.. [1] These are rate limited by default (see default mask above)
1512
1513icmp_ignore_bogus_error_responses - BOOLEAN
1514	Some routers violate RFC1122 by sending bogus responses to broadcast
1515	frames.  Such violations are normally logged via a kernel warning.
1516	If this is set to TRUE, the kernel will not give such warnings, which
1517	will avoid log file clutter.
1518
1519	Default: 1
1520
1521icmp_errors_use_inbound_ifaddr - BOOLEAN
1522
1523	If zero, icmp error messages are sent with the primary address of
1524	the exiting interface.
1525
1526	If non-zero, the message will be sent with the primary address of
1527	the interface that received the packet that caused the icmp error.
1528	This is the behaviour many network administrators will expect from
1529	a router. And it can make debugging complicated network layouts
1530	much easier.
1531
1532	Note that if no primary address exists for the interface selected,
1533	then the primary address of the first non-loopback interface that
1534	has one will be used regardless of this setting.
1535
1536	Default: 0
1537
1538igmp_max_memberships - INTEGER
1539	Change the maximum number of multicast groups we can subscribe to.
1540	Default: 20
1541
1542	Theoretical maximum value is bounded by having to send a membership
1543	report in a single datagram (i.e. the report can't span multiple
1544	datagrams, or risk confusing the switch and leaving groups you don't
1545	intend to).
1546
1547	The number of supported groups 'M' is bounded by the number of group
1548	report entries you can fit into a single datagram of 65535 bytes.
1549
1550	M = 65536-sizeof (ip header)/(sizeof(Group record))
1551
1552	Group records are variable length, with a minimum of 12 bytes.
1553	So net.ipv4.igmp_max_memberships should not be set higher than:
1554
1555	(65536-24) / 12 = 5459
1556
1557	The value 5459 assumes no IP header options, so in practice
1558	this number may be lower.
1559
1560igmp_max_msf - INTEGER
1561	Maximum number of addresses allowed in the source filter list for a
1562	multicast group.
1563
1564	Default: 10
1565
1566igmp_qrv - INTEGER
1567	Controls the IGMP query robustness variable (see RFC2236 8.1).
1568
1569	Default: 2 (as specified by RFC2236 8.1)
1570
1571	Minimum: 1 (as specified by RFC6636 4.5)
1572
1573force_igmp_version - INTEGER
1574	- 0 - (default) No enforcement of a IGMP version, IGMPv1/v2 fallback
1575	  allowed. Will back to IGMPv3 mode again if all IGMPv1/v2 Querier
1576	  Present timer expires.
1577	- 1 - Enforce to use IGMP version 1. Will also reply IGMPv1 report if
1578	  receive IGMPv2/v3 query.
1579	- 2 - Enforce to use IGMP version 2. Will fallback to IGMPv1 if receive
1580	  IGMPv1 query message. Will reply report if receive IGMPv3 query.
1581	- 3 - Enforce to use IGMP version 3. The same react with default 0.
1582
1583	.. note::
1584
1585	   this is not the same with force_mld_version because IGMPv3 RFC3376
1586	   Security Considerations does not have clear description that we could
1587	   ignore other version messages completely as MLDv2 RFC3810. So make
1588	   this value as default 0 is recommended.
1589
1590``conf/interface/*``
1591	changes special settings per interface (where
1592	interface" is the name of your network interface)
1593
1594``conf/all/*``
1595	  is special, changes the settings for all interfaces
1596
1597log_martians - BOOLEAN
1598	Log packets with impossible addresses to kernel log.
1599	log_martians for the interface will be enabled if at least one of
1600	conf/{all,interface}/log_martians is set to TRUE,
1601	it will be disabled otherwise
1602
1603accept_redirects - BOOLEAN
1604	Accept ICMP redirect messages.
1605	accept_redirects for the interface will be enabled if:
1606
1607	- both conf/{all,interface}/accept_redirects are TRUE in the case
1608	  forwarding for the interface is enabled
1609
1610	or
1611
1612	- at least one of conf/{all,interface}/accept_redirects is TRUE in the
1613	  case forwarding for the interface is disabled
1614
1615	accept_redirects for the interface will be disabled otherwise
1616
1617	default:
1618
1619		- TRUE (host)
1620		- FALSE (router)
1621
1622forwarding - BOOLEAN
1623	Enable IP forwarding on this interface.  This controls whether packets
1624	received _on_ this interface can be forwarded.
1625
1626mc_forwarding - BOOLEAN
1627	Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
1628	and a multicast routing daemon is required.
1629	conf/all/mc_forwarding must also be set to TRUE to enable multicast
1630	routing	for the interface
1631
1632medium_id - INTEGER
1633	Integer value used to differentiate the devices by the medium they
1634	are attached to. Two devices can have different id values when
1635	the broadcast packets are received only on one of them.
1636	The default value 0 means that the device is the only interface
1637	to its medium, value of -1 means that medium is not known.
1638
1639	Currently, it is used to change the proxy_arp behavior:
1640	the proxy_arp feature is enabled for packets forwarded between
1641	two devices attached to different media.
1642
1643proxy_arp - BOOLEAN
1644	Do proxy arp.
1645
1646	proxy_arp for the interface will be enabled if at least one of
1647	conf/{all,interface}/proxy_arp is set to TRUE,
1648	it will be disabled otherwise
1649
1650proxy_arp_pvlan - BOOLEAN
1651	Private VLAN proxy arp.
1652
1653	Basically allow proxy arp replies back to the same interface
1654	(from which the ARP request/solicitation was received).
1655
1656	This is done to support (ethernet) switch features, like RFC
1657	3069, where the individual ports are NOT allowed to
1658	communicate with each other, but they are allowed to talk to
1659	the upstream router.  As described in RFC 3069, it is possible
1660	to allow these hosts to communicate through the upstream
1661	router by proxy_arp'ing. Don't need to be used together with
1662	proxy_arp.
1663
1664	This technology is known by different names:
1665
1666	  In RFC 3069 it is called VLAN Aggregation.
1667	  Cisco and Allied Telesyn call it Private VLAN.
1668	  Hewlett-Packard call it Source-Port filtering or port-isolation.
1669	  Ericsson call it MAC-Forced Forwarding (RFC Draft).
1670
1671proxy_delay - INTEGER
1672	Delay proxy response.
1673
1674	Delay response to a neighbor solicitation when proxy_arp
1675	or proxy_ndp is enabled. A random value between [0, proxy_delay)
1676	will be chosen, setting to zero means reply with no delay.
1677	Value in jiffies. Defaults to 80.
1678
1679shared_media - BOOLEAN
1680	Send(router) or accept(host) RFC1620 shared media redirects.
1681	Overrides secure_redirects.
1682
1683	shared_media for the interface will be enabled if at least one of
1684	conf/{all,interface}/shared_media is set to TRUE,
1685	it will be disabled otherwise
1686
1687	default TRUE
1688
1689secure_redirects - BOOLEAN
1690	Accept ICMP redirect messages only to gateways listed in the
1691	interface's current gateway list. Even if disabled, RFC1122 redirect
1692	rules still apply.
1693
1694	Overridden by shared_media.
1695
1696	secure_redirects for the interface will be enabled if at least one of
1697	conf/{all,interface}/secure_redirects is set to TRUE,
1698	it will be disabled otherwise
1699
1700	default TRUE
1701
1702send_redirects - BOOLEAN
1703	Send redirects, if router.
1704
1705	send_redirects for the interface will be enabled if at least one of
1706	conf/{all,interface}/send_redirects is set to TRUE,
1707	it will be disabled otherwise
1708
1709	Default: TRUE
1710
1711bootp_relay - BOOLEAN
1712	Accept packets with source address 0.b.c.d destined
1713	not to this host as local ones. It is supposed, that
1714	BOOTP relay daemon will catch and forward such packets.
1715	conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
1716	for the interface
1717
1718	default FALSE
1719
1720	Not Implemented Yet.
1721
1722accept_source_route - BOOLEAN
1723	Accept packets with SRR option.
1724	conf/all/accept_source_route must also be set to TRUE to accept packets
1725	with SRR option on the interface
1726
1727	default
1728
1729		- TRUE (router)
1730		- FALSE (host)
1731
1732accept_local - BOOLEAN
1733	Accept packets with local source addresses. In combination with
1734	suitable routing, this can be used to direct packets between two
1735	local interfaces over the wire and have them accepted properly.
1736	default FALSE
1737
1738route_localnet - BOOLEAN
1739	Do not consider loopback addresses as martian source or destination
1740	while routing. This enables the use of 127/8 for local routing purposes.
1741
1742	default FALSE
1743
1744rp_filter - INTEGER
1745	- 0 - No source validation.
1746	- 1 - Strict mode as defined in RFC3704 Strict Reverse Path
1747	  Each incoming packet is tested against the FIB and if the interface
1748	  is not the best reverse path the packet check will fail.
1749	  By default failed packets are discarded.
1750	- 2 - Loose mode as defined in RFC3704 Loose Reverse Path
1751	  Each incoming packet's source address is also tested against the FIB
1752	  and if the source address is not reachable via any interface
1753	  the packet check will fail.
1754
1755	Current recommended practice in RFC3704 is to enable strict mode
1756	to prevent IP spoofing from DDos attacks. If using asymmetric routing
1757	or other complicated routing, then loose mode is recommended.
1758
1759	The max value from conf/{all,interface}/rp_filter is used
1760	when doing source validation on the {interface}.
1761
1762	Default value is 0. Note that some distributions enable it
1763	in startup scripts.
1764
1765src_valid_mark - BOOLEAN
1766	- 0 - The fwmark of the packet is not included in reverse path
1767	  route lookup.  This allows for asymmetric routing configurations
1768	  utilizing the fwmark in only one direction, e.g., transparent
1769	  proxying.
1770
1771	- 1 - The fwmark of the packet is included in reverse path route
1772	  lookup.  This permits rp_filter to function when the fwmark is
1773	  used for routing traffic in both directions.
1774
1775	This setting also affects the utilization of fmwark when
1776	performing source address selection for ICMP replies, or
1777	determining addresses stored for the IPOPT_TS_TSANDADDR and
1778	IPOPT_RR IP options.
1779
1780	The max value from conf/{all,interface}/src_valid_mark is used.
1781
1782	Default value is 0.
1783
1784arp_filter - BOOLEAN
1785	- 1 - Allows you to have multiple network interfaces on the same
1786	  subnet, and have the ARPs for each interface be answered
1787	  based on whether or not the kernel would route a packet from
1788	  the ARP'd IP out that interface (therefore you must use source
1789	  based routing for this to work). In other words it allows control
1790	  of which cards (usually 1) will respond to an arp request.
1791
1792	- 0 - (default) The kernel can respond to arp requests with addresses
1793	  from other interfaces. This may seem wrong but it usually makes
1794	  sense, because it increases the chance of successful communication.
1795	  IP addresses are owned by the complete host on Linux, not by
1796	  particular interfaces. Only for more complex setups like load-
1797	  balancing, does this behaviour cause problems.
1798
1799	arp_filter for the interface will be enabled if at least one of
1800	conf/{all,interface}/arp_filter is set to TRUE,
1801	it will be disabled otherwise
1802
1803arp_announce - INTEGER
1804	Define different restriction levels for announcing the local
1805	source IP address from IP packets in ARP requests sent on
1806	interface:
1807
1808	- 0 - (default) Use any local address, configured on any interface
1809	- 1 - Try to avoid local addresses that are not in the target's
1810	  subnet for this interface. This mode is useful when target
1811	  hosts reachable via this interface require the source IP
1812	  address in ARP requests to be part of their logical network
1813	  configured on the receiving interface. When we generate the
1814	  request we will check all our subnets that include the
1815	  target IP and will preserve the source address if it is from
1816	  such subnet. If there is no such subnet we select source
1817	  address according to the rules for level 2.
1818	- 2 - Always use the best local address for this target.
1819	  In this mode we ignore the source address in the IP packet
1820	  and try to select local address that we prefer for talks with
1821	  the target host. Such local address is selected by looking
1822	  for primary IP addresses on all our subnets on the outgoing
1823	  interface that include the target IP address. If no suitable
1824	  local address is found we select the first local address
1825	  we have on the outgoing interface or on all other interfaces,
1826	  with the hope we will receive reply for our request and
1827	  even sometimes no matter the source IP address we announce.
1828
1829	The max value from conf/{all,interface}/arp_announce is used.
1830
1831	Increasing the restriction level gives more chance for
1832	receiving answer from the resolved target while decreasing
1833	the level announces more valid sender's information.
1834
1835arp_ignore - INTEGER
1836	Define different modes for sending replies in response to
1837	received ARP requests that resolve local target IP addresses:
1838
1839	- 0 - (default): reply for any local target IP address, configured
1840	  on any interface
1841	- 1 - reply only if the target IP address is local address
1842	  configured on the incoming interface
1843	- 2 - reply only if the target IP address is local address
1844	  configured on the incoming interface and both with the
1845	  sender's IP address are part from same subnet on this interface
1846	- 3 - do not reply for local addresses configured with scope host,
1847	  only resolutions for global and link addresses are replied
1848	- 4-7 - reserved
1849	- 8 - do not reply for all local addresses
1850
1851	The max value from conf/{all,interface}/arp_ignore is used
1852	when ARP request is received on the {interface}
1853
1854arp_notify - BOOLEAN
1855	Define mode for notification of address and device changes.
1856
1857	 ==  ==========================================================
1858	  0  (default): do nothing
1859	  1  Generate gratuitous arp requests when device is brought up
1860	     or hardware address changes.
1861	 ==  ==========================================================
1862
1863arp_accept - INTEGER
1864	Define behavior for accepting gratuitous ARP (garp) frames from devices
1865	that are not already present in the ARP table:
1866
1867	- 0 - don't create new entries in the ARP table
1868	- 1 - create new entries in the ARP table
1869	- 2 - create new entries only if the source IP address is in the same
1870	  subnet as an address configured on the interface that received the
1871	  garp message.
1872
1873	Both replies and requests type gratuitous arp will trigger the
1874	ARP table to be updated, if this setting is on.
1875
1876	If the ARP table already contains the IP address of the
1877	gratuitous arp frame, the arp table will be updated regardless
1878	if this setting is on or off.
1879
1880arp_evict_nocarrier - BOOLEAN
1881	Clears the ARP cache on NOCARRIER events. This option is important for
1882	wireless devices where the ARP cache should not be cleared when roaming
1883	between access points on the same network. In most cases this should
1884	remain as the default (1).
1885
1886	- 1 - (default): Clear the ARP cache on NOCARRIER events
1887	- 0 - Do not clear ARP cache on NOCARRIER events
1888
1889mcast_solicit - INTEGER
1890	The maximum number of multicast probes in INCOMPLETE state,
1891	when the associated hardware address is unknown.  Defaults
1892	to 3.
1893
1894ucast_solicit - INTEGER
1895	The maximum number of unicast probes in PROBE state, when
1896	the hardware address is being reconfirmed.  Defaults to 3.
1897
1898app_solicit - INTEGER
1899	The maximum number of probes to send to the user space ARP daemon
1900	via netlink before dropping back to multicast probes (see
1901	mcast_resolicit).  Defaults to 0.
1902
1903mcast_resolicit - INTEGER
1904	The maximum number of multicast probes after unicast and
1905	app probes in PROBE state.  Defaults to 0.
1906
1907disable_policy - BOOLEAN
1908	Disable IPSEC policy (SPD) for this interface
1909
1910disable_xfrm - BOOLEAN
1911	Disable IPSEC encryption on this interface, whatever the policy
1912
1913igmpv2_unsolicited_report_interval - INTEGER
1914	The interval in milliseconds in which the next unsolicited
1915	IGMPv1 or IGMPv2 report retransmit will take place.
1916
1917	Default: 10000 (10 seconds)
1918
1919igmpv3_unsolicited_report_interval - INTEGER
1920	The interval in milliseconds in which the next unsolicited
1921	IGMPv3 report retransmit will take place.
1922
1923	Default: 1000 (1 seconds)
1924
1925ignore_routes_with_linkdown - BOOLEAN
1926        Ignore routes whose link is down when performing a FIB lookup.
1927
1928promote_secondaries - BOOLEAN
1929	When a primary IP address is removed from this interface
1930	promote a corresponding secondary IP address instead of
1931	removing all the corresponding secondary IP addresses.
1932
1933drop_unicast_in_l2_multicast - BOOLEAN
1934	Drop any unicast IP packets that are received in link-layer
1935	multicast (or broadcast) frames.
1936
1937	This behavior (for multicast) is actually a SHOULD in RFC
1938	1122, but is disabled by default for compatibility reasons.
1939
1940	Default: off (0)
1941
1942drop_gratuitous_arp - BOOLEAN
1943	Drop all gratuitous ARP frames, for example if there's a known
1944	good ARP proxy on the network and such frames need not be used
1945	(or in the case of 802.11, must not be used to prevent attacks.)
1946
1947	Default: off (0)
1948
1949
1950tag - INTEGER
1951	Allows you to write a number, which can be used as required.
1952
1953	Default value is 0.
1954
1955xfrm4_gc_thresh - INTEGER
1956	(Obsolete since linux-4.14)
1957	The threshold at which we will start garbage collecting for IPv4
1958	destination cache entries.  At twice this value the system will
1959	refuse new allocations.
1960
1961igmp_link_local_mcast_reports - BOOLEAN
1962	Enable IGMP reports for link local multicast groups in the
1963	224.0.0.X range.
1964
1965	Default TRUE
1966
1967Alexey Kuznetsov.
1968kuznet@ms2.inr.ac.ru
1969
1970Updated by:
1971
1972- Andi Kleen
1973  ak@muc.de
1974- Nicolas Delon
1975  delon.nicolas@wanadoo.fr
1976
1977
1978
1979
1980/proc/sys/net/ipv6/* Variables
1981==============================
1982
1983IPv6 has no global variables such as tcp_*.  tcp_* settings under ipv4/ also
1984apply to IPv6 [XXX?].
1985
1986bindv6only - BOOLEAN
1987	Default value for IPV6_V6ONLY socket option,
1988	which restricts use of the IPv6 socket to IPv6 communication
1989	only.
1990
1991		- TRUE: disable IPv4-mapped address feature
1992		- FALSE: enable IPv4-mapped address feature
1993
1994	Default: FALSE (as specified in RFC3493)
1995
1996flowlabel_consistency - BOOLEAN
1997	Protect the consistency (and unicity) of flow label.
1998	You have to disable it to use IPV6_FL_F_REFLECT flag on the
1999	flow label manager.
2000
2001	- TRUE: enabled
2002	- FALSE: disabled
2003
2004	Default: TRUE
2005
2006auto_flowlabels - INTEGER
2007	Automatically generate flow labels based on a flow hash of the
2008	packet. This allows intermediate devices, such as routers, to
2009	identify packet flows for mechanisms like Equal Cost Multipath
2010	Routing (see RFC 6438).
2011
2012	=  ===========================================================
2013	0  automatic flow labels are completely disabled
2014	1  automatic flow labels are enabled by default, they can be
2015	   disabled on a per socket basis using the IPV6_AUTOFLOWLABEL
2016	   socket option
2017	2  automatic flow labels are allowed, they may be enabled on a
2018	   per socket basis using the IPV6_AUTOFLOWLABEL socket option
2019	3  automatic flow labels are enabled and enforced, they cannot
2020	   be disabled by the socket option
2021	=  ===========================================================
2022
2023	Default: 1
2024
2025flowlabel_state_ranges - BOOLEAN
2026	Split the flow label number space into two ranges. 0-0x7FFFF is
2027	reserved for the IPv6 flow manager facility, 0x80000-0xFFFFF
2028	is reserved for stateless flow labels as described in RFC6437.
2029
2030	- TRUE: enabled
2031	- FALSE: disabled
2032
2033	Default: true
2034
2035flowlabel_reflect - INTEGER
2036	Control flow label reflection. Needed for Path MTU
2037	Discovery to work with Equal Cost Multipath Routing in anycast
2038	environments. See RFC 7690 and:
2039	https://tools.ietf.org/html/draft-wang-6man-flow-label-reflection-01
2040
2041	This is a bitmask.
2042
2043	- 1: enabled for established flows
2044
2045	  Note that this prevents automatic flowlabel changes, as done
2046	  in "tcp: change IPv6 flow-label upon receiving spurious retransmission"
2047	  and "tcp: Change txhash on every SYN and RTO retransmit"
2048
2049	- 2: enabled for TCP RESET packets (no active listener)
2050	  If set, a RST packet sent in response to a SYN packet on a closed
2051	  port will reflect the incoming flow label.
2052
2053	- 4: enabled for ICMPv6 echo reply messages.
2054
2055	Default: 0
2056
2057fib_multipath_hash_policy - INTEGER
2058	Controls which hash policy to use for multipath routes.
2059
2060	Default: 0 (Layer 3)
2061
2062	Possible values:
2063
2064	- 0 - Layer 3 (source and destination addresses plus flow label)
2065	- 1 - Layer 4 (standard 5-tuple)
2066	- 2 - Layer 3 or inner Layer 3 if present
2067	- 3 - Custom multipath hash. Fields used for multipath hash calculation
2068	  are determined by fib_multipath_hash_fields sysctl
2069
2070fib_multipath_hash_fields - UNSIGNED INTEGER
2071	When fib_multipath_hash_policy is set to 3 (custom multipath hash), the
2072	fields used for multipath hash calculation are determined by this
2073	sysctl.
2074
2075	This value is a bitmask which enables various fields for multipath hash
2076	calculation.
2077
2078	Possible fields are:
2079
2080	====== ============================
2081	0x0001 Source IP address
2082	0x0002 Destination IP address
2083	0x0004 IP protocol
2084	0x0008 Flow Label
2085	0x0010 Source port
2086	0x0020 Destination port
2087	0x0040 Inner source IP address
2088	0x0080 Inner destination IP address
2089	0x0100 Inner IP protocol
2090	0x0200 Inner Flow Label
2091	0x0400 Inner source port
2092	0x0800 Inner destination port
2093	====== ============================
2094
2095	Default: 0x0007 (source IP, destination IP and IP protocol)
2096
2097anycast_src_echo_reply - BOOLEAN
2098	Controls the use of anycast addresses as source addresses for ICMPv6
2099	echo reply
2100
2101	- TRUE:  enabled
2102	- FALSE: disabled
2103
2104	Default: FALSE
2105
2106idgen_delay - INTEGER
2107	Controls the delay in seconds after which time to retry
2108	privacy stable address generation if a DAD conflict is
2109	detected.
2110
2111	Default: 1 (as specified in RFC7217)
2112
2113idgen_retries - INTEGER
2114	Controls the number of retries to generate a stable privacy
2115	address if a DAD conflict is detected.
2116
2117	Default: 3 (as specified in RFC7217)
2118
2119mld_qrv - INTEGER
2120	Controls the MLD query robustness variable (see RFC3810 9.1).
2121
2122	Default: 2 (as specified by RFC3810 9.1)
2123
2124	Minimum: 1 (as specified by RFC6636 4.5)
2125
2126max_dst_opts_number - INTEGER
2127	Maximum number of non-padding TLVs allowed in a Destination
2128	options extension header. If this value is less than zero
2129	then unknown options are disallowed and the number of known
2130	TLVs allowed is the absolute value of this number.
2131
2132	Default: 8
2133
2134max_hbh_opts_number - INTEGER
2135	Maximum number of non-padding TLVs allowed in a Hop-by-Hop
2136	options extension header. If this value is less than zero
2137	then unknown options are disallowed and the number of known
2138	TLVs allowed is the absolute value of this number.
2139
2140	Default: 8
2141
2142max_dst_opts_length - INTEGER
2143	Maximum length allowed for a Destination options extension
2144	header.
2145
2146	Default: INT_MAX (unlimited)
2147
2148max_hbh_length - INTEGER
2149	Maximum length allowed for a Hop-by-Hop options extension
2150	header.
2151
2152	Default: INT_MAX (unlimited)
2153
2154skip_notify_on_dev_down - BOOLEAN
2155	Controls whether an RTM_DELROUTE message is generated for routes
2156	removed when a device is taken down or deleted. IPv4 does not
2157	generate this message; IPv6 does by default. Setting this sysctl
2158	to true skips the message, making IPv4 and IPv6 on par in relying
2159	on userspace caches to track link events and evict routes.
2160
2161	Default: false (generate message)
2162
2163nexthop_compat_mode - BOOLEAN
2164	New nexthop API provides a means for managing nexthops independent of
2165	prefixes. Backwards compatibility with old route format is enabled by
2166	default which means route dumps and notifications contain the new
2167	nexthop attribute but also the full, expanded nexthop definition.
2168	Further, updates or deletes of a nexthop configuration generate route
2169	notifications for each fib entry using the nexthop. Once a system
2170	understands the new API, this sysctl can be disabled to achieve full
2171	performance benefits of the new API by disabling the nexthop expansion
2172	and extraneous notifications.
2173
2174	Note that as a backward-compatible mode, dumping of modern features
2175	might be incomplete or wrong. For example, resilient groups will not be
2176	shown as such, but rather as just a list of next hops. Also weights that
2177	do not fit into 8 bits will show incorrectly.
2178
2179	Default: true (backward compat mode)
2180
2181fib_notify_on_flag_change - INTEGER
2182        Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
2183        RTM_F_TRAP/RTM_F_OFFLOAD_FAILED flags are changed.
2184
2185        After installing a route to the kernel, user space receives an
2186        acknowledgment, which means the route was installed in the kernel,
2187        but not necessarily in hardware.
2188        It is also possible for a route already installed in hardware to change
2189        its action and therefore its flags. For example, a host route that is
2190        trapping packets can be "promoted" to perform decapsulation following
2191        the installation of an IPinIP/VXLAN tunnel.
2192        The notifications will indicate to user-space the state of the route.
2193
2194        Default: 0 (Do not emit notifications.)
2195
2196        Possible values:
2197
2198        - 0 - Do not emit notifications.
2199        - 1 - Emit notifications.
2200        - 2 - Emit notifications only for RTM_F_OFFLOAD_FAILED flag change.
2201
2202ioam6_id - INTEGER
2203        Define the IOAM id of this node. Uses only 24 bits out of 32 in total.
2204
2205        Min: 0
2206        Max: 0xFFFFFF
2207
2208        Default: 0xFFFFFF
2209
2210ioam6_id_wide - LONG INTEGER
2211        Define the wide IOAM id of this node. Uses only 56 bits out of 64 in
2212        total. Can be different from ioam6_id.
2213
2214        Min: 0
2215        Max: 0xFFFFFFFFFFFFFF
2216
2217        Default: 0xFFFFFFFFFFFFFF
2218
2219IPv6 Fragmentation:
2220
2221ip6frag_high_thresh - INTEGER
2222	Maximum memory used to reassemble IPv6 fragments. When
2223	ip6frag_high_thresh bytes of memory is allocated for this purpose,
2224	the fragment handler will toss packets until ip6frag_low_thresh
2225	is reached.
2226
2227ip6frag_low_thresh - INTEGER
2228	See ip6frag_high_thresh
2229
2230ip6frag_time - INTEGER
2231	Time in seconds to keep an IPv6 fragment in memory.
2232
2233``conf/default/*``:
2234	Change the interface-specific default settings.
2235
2236	These settings would be used during creating new interfaces.
2237
2238
2239``conf/all/*``:
2240	Change all the interface-specific settings.
2241
2242	[XXX:  Other special features than forwarding?]
2243
2244conf/all/disable_ipv6 - BOOLEAN
2245	Changing this value is same as changing ``conf/default/disable_ipv6``
2246	setting and also all per-interface ``disable_ipv6`` settings to the same
2247	value.
2248
2249	Reading this value does not have any particular meaning. It does not say
2250	whether IPv6 support is enabled or disabled. Returned value can be 1
2251	also in the case when some interface has ``disable_ipv6`` set to 0 and
2252	has configured IPv6 addresses.
2253
2254conf/all/forwarding - BOOLEAN
2255	Enable global IPv6 forwarding between all interfaces.
2256
2257	IPv4 and IPv6 work differently here; e.g. netfilter must be used
2258	to control which interfaces may forward packets and which not.
2259
2260	This also sets all interfaces' Host/Router setting
2261	'forwarding' to the specified value.  See below for details.
2262
2263	This referred to as global forwarding.
2264
2265proxy_ndp - BOOLEAN
2266	Do proxy ndp.
2267
2268fwmark_reflect - BOOLEAN
2269	Controls the fwmark of kernel-generated IPv6 reply packets that are not
2270	associated with a socket for example, TCP RSTs or ICMPv6 echo replies).
2271	If unset, these packets have a fwmark of zero. If set, they have the
2272	fwmark of the packet they are replying to.
2273
2274	Default: 0
2275
2276``conf/interface/*``:
2277	Change special settings per interface.
2278
2279	The functional behaviour for certain settings is different
2280	depending on whether local forwarding is enabled or not.
2281
2282accept_ra - INTEGER
2283	Accept Router Advertisements; autoconfigure using them.
2284
2285	It also determines whether or not to transmit Router
2286	Solicitations. If and only if the functional setting is to
2287	accept Router Advertisements, Router Solicitations will be
2288	transmitted.
2289
2290	Possible values are:
2291
2292		==  ===========================================================
2293		 0  Do not accept Router Advertisements.
2294		 1  Accept Router Advertisements if forwarding is disabled.
2295		 2  Overrule forwarding behaviour. Accept Router Advertisements
2296		    even if forwarding is enabled.
2297		==  ===========================================================
2298
2299	Functional default:
2300
2301		- enabled if local forwarding is disabled.
2302		- disabled if local forwarding is enabled.
2303
2304accept_ra_defrtr - BOOLEAN
2305	Learn default router in Router Advertisement.
2306
2307	Functional default:
2308
2309		- enabled if accept_ra is enabled.
2310		- disabled if accept_ra is disabled.
2311
2312ra_defrtr_metric - UNSIGNED INTEGER
2313	Route metric for default route learned in Router Advertisement. This value
2314	will be assigned as metric for the default route learned via IPv6 Router
2315	Advertisement. Takes affect only if accept_ra_defrtr is enabled.
2316
2317	Possible values:
2318		1 to 0xFFFFFFFF
2319
2320		Default: IP6_RT_PRIO_USER i.e. 1024.
2321
2322accept_ra_from_local - BOOLEAN
2323	Accept RA with source-address that is found on local machine
2324	if the RA is otherwise proper and able to be accepted.
2325
2326	Default is to NOT accept these as it may be an un-intended
2327	network loop.
2328
2329	Functional default:
2330
2331	   - enabled if accept_ra_from_local is enabled
2332	     on a specific interface.
2333	   - disabled if accept_ra_from_local is disabled
2334	     on a specific interface.
2335
2336accept_ra_min_hop_limit - INTEGER
2337	Minimum hop limit Information in Router Advertisement.
2338
2339	Hop limit Information in Router Advertisement less than this
2340	variable shall be ignored.
2341
2342	Default: 1
2343
2344accept_ra_min_lft - INTEGER
2345	Minimum acceptable lifetime value in Router Advertisement.
2346
2347	RA sections with a lifetime less than this value shall be
2348	ignored. Zero lifetimes stay unaffected.
2349
2350	Default: 0
2351
2352accept_ra_pinfo - BOOLEAN
2353	Learn Prefix Information in Router Advertisement.
2354
2355	Functional default:
2356
2357		- enabled if accept_ra is enabled.
2358		- disabled if accept_ra is disabled.
2359
2360ra_honor_pio_life - BOOLEAN
2361	Whether to use RFC4862 Section 5.5.3e to determine the valid
2362	lifetime of an address matching a prefix sent in a Router
2363	Advertisement Prefix Information Option.
2364
2365	- If enabled, the PIO valid lifetime will always be honored.
2366	- If disabled, RFC4862 section 5.5.3e is used to determine
2367	  the valid lifetime of the address.
2368
2369	Default: 0 (disabled)
2370
2371ra_honor_pio_pflag - BOOLEAN
2372	The Prefix Information Option P-flag indicates the network can
2373	allocate a unique IPv6 prefix per client using DHCPv6-PD.
2374	This sysctl can be enabled when a userspace DHCPv6-PD client
2375	is running to cause the P-flag to take effect: i.e. the
2376	P-flag suppresses any effects of the A-flag within the same
2377	PIO. For a given PIO, P=1 and A=1 is treated as A=0.
2378
2379	- If disabled, the P-flag is ignored.
2380	- If enabled, the P-flag will disable SLAAC autoconfiguration
2381	  for the given Prefix Information Option.
2382
2383	Default: 0 (disabled)
2384
2385accept_ra_rt_info_min_plen - INTEGER
2386	Minimum prefix length of Route Information in RA.
2387
2388	Route Information w/ prefix smaller than this variable shall
2389	be ignored.
2390
2391	Functional default:
2392
2393		* 0 if accept_ra_rtr_pref is enabled.
2394		* -1 if accept_ra_rtr_pref is disabled.
2395
2396accept_ra_rt_info_max_plen - INTEGER
2397	Maximum prefix length of Route Information in RA.
2398
2399	Route Information w/ prefix larger than this variable shall
2400	be ignored.
2401
2402	Functional default:
2403
2404		* 0 if accept_ra_rtr_pref is enabled.
2405		* -1 if accept_ra_rtr_pref is disabled.
2406
2407accept_ra_rtr_pref - BOOLEAN
2408	Accept Router Preference in RA.
2409
2410	Functional default:
2411
2412		- enabled if accept_ra is enabled.
2413		- disabled if accept_ra is disabled.
2414
2415accept_ra_mtu - BOOLEAN
2416	Apply the MTU value specified in RA option 5 (RFC4861). If
2417	disabled, the MTU specified in the RA will be ignored.
2418
2419	Functional default:
2420
2421		- enabled if accept_ra is enabled.
2422		- disabled if accept_ra is disabled.
2423
2424accept_redirects - BOOLEAN
2425	Accept Redirects.
2426
2427	Functional default:
2428
2429		- enabled if local forwarding is disabled.
2430		- disabled if local forwarding is enabled.
2431
2432accept_source_route - INTEGER
2433	Accept source routing (routing extension header).
2434
2435	- >= 0: Accept only routing header type 2.
2436	- < 0: Do not accept routing header.
2437
2438	Default: 0
2439
2440autoconf - BOOLEAN
2441	Autoconfigure addresses using Prefix Information in Router
2442	Advertisements.
2443
2444	Functional default:
2445
2446		- enabled if accept_ra_pinfo is enabled.
2447		- disabled if accept_ra_pinfo is disabled.
2448
2449dad_transmits - INTEGER
2450	The amount of Duplicate Address Detection probes to send.
2451
2452	Default: 1
2453
2454forwarding - INTEGER
2455	Configure interface-specific Host/Router behaviour.
2456
2457	.. note::
2458
2459	   It is recommended to have the same setting on all
2460	   interfaces; mixed router/host scenarios are rather uncommon.
2461
2462	Possible values are:
2463
2464		- 0 Forwarding disabled
2465		- 1 Forwarding enabled
2466
2467	**FALSE (0)**:
2468
2469	By default, Host behaviour is assumed.  This means:
2470
2471	1. IsRouter flag is not set in Neighbour Advertisements.
2472	2. If accept_ra is TRUE (default), transmit Router
2473	   Solicitations.
2474	3. If accept_ra is TRUE (default), accept Router
2475	   Advertisements (and do autoconfiguration).
2476	4. If accept_redirects is TRUE (default), accept Redirects.
2477
2478	**TRUE (1)**:
2479
2480	If local forwarding is enabled, Router behaviour is assumed.
2481	This means exactly the reverse from the above:
2482
2483	1. IsRouter flag is set in Neighbour Advertisements.
2484	2. Router Solicitations are not sent unless accept_ra is 2.
2485	3. Router Advertisements are ignored unless accept_ra is 2.
2486	4. Redirects are ignored.
2487
2488	Default: 0 (disabled) if global forwarding is disabled (default),
2489	otherwise 1 (enabled).
2490
2491hop_limit - INTEGER
2492	Default Hop Limit to set.
2493
2494	Default: 64
2495
2496mtu - INTEGER
2497	Default Maximum Transfer Unit
2498
2499	Default: 1280 (IPv6 required minimum)
2500
2501ip_nonlocal_bind - BOOLEAN
2502	If set, allows processes to bind() to non-local IPv6 addresses,
2503	which can be quite useful - but may break some applications.
2504
2505	Default: 0
2506
2507router_probe_interval - INTEGER
2508	Minimum interval (in seconds) between Router Probing described
2509	in RFC4191.
2510
2511	Default: 60
2512
2513router_solicitation_delay - INTEGER
2514	Number of seconds to wait after interface is brought up
2515	before sending Router Solicitations.
2516
2517	Default: 1
2518
2519router_solicitation_interval - INTEGER
2520	Number of seconds to wait between Router Solicitations.
2521
2522	Default: 4
2523
2524router_solicitations - INTEGER
2525	Number of Router Solicitations to send until assuming no
2526	routers are present.
2527
2528	Default: 3
2529
2530use_oif_addrs_only - BOOLEAN
2531	When enabled, the candidate source addresses for destinations
2532	routed via this interface are restricted to the set of addresses
2533	configured on this interface (vis. RFC 6724, section 4).
2534
2535	Default: false
2536
2537use_tempaddr - INTEGER
2538	Preference for Privacy Extensions (RFC3041).
2539
2540	  * <= 0 : disable Privacy Extensions
2541	  * == 1 : enable Privacy Extensions, but prefer public
2542	    addresses over temporary addresses.
2543	  * >  1 : enable Privacy Extensions and prefer temporary
2544	    addresses over public addresses.
2545
2546	Default:
2547
2548		* 0 (for most devices)
2549		* -1 (for point-to-point devices and loopback devices)
2550
2551temp_valid_lft - INTEGER
2552	valid lifetime (in seconds) for temporary addresses. If less than the
2553	minimum required lifetime (typically 5-7 seconds), temporary addresses
2554	will not be created.
2555
2556	Default: 172800 (2 days)
2557
2558temp_prefered_lft - INTEGER
2559	Preferred lifetime (in seconds) for temporary addresses. If
2560	temp_prefered_lft is less than the minimum required lifetime (typically
2561	5-7 seconds), the preferred lifetime is the minimum required. If
2562	temp_prefered_lft is greater than temp_valid_lft, the preferred lifetime
2563	is temp_valid_lft.
2564
2565	Default: 86400 (1 day)
2566
2567keep_addr_on_down - INTEGER
2568	Keep all IPv6 addresses on an interface down event. If set static
2569	global addresses with no expiration time are not flushed.
2570
2571	*   >0 : enabled
2572	*    0 : system default
2573	*   <0 : disabled
2574
2575	Default: 0 (addresses are removed)
2576
2577max_desync_factor - INTEGER
2578	Maximum value for DESYNC_FACTOR, which is a random value
2579	that ensures that clients don't synchronize with each
2580	other and generate new addresses at exactly the same time.
2581	value is in seconds.
2582
2583	Default: 600
2584
2585regen_min_advance - INTEGER
2586	How far in advance (in seconds), at minimum, to create a new temporary
2587	address before the current one is deprecated. This value is added to
2588	the amount of time that may be required for duplicate address detection
2589	to determine when to create a new address. Linux permits setting this
2590	value to less than the default of 2 seconds, but a value less than 2
2591	does not conform to RFC 8981.
2592
2593	Default: 2
2594
2595regen_max_retry - INTEGER
2596	Number of attempts before give up attempting to generate
2597	valid temporary addresses.
2598
2599	Default: 5
2600
2601max_addresses - INTEGER
2602	Maximum number of autoconfigured addresses per interface.  Setting
2603	to zero disables the limitation.  It is not recommended to set this
2604	value too large (or to zero) because it would be an easy way to
2605	crash the kernel by allowing too many addresses to be created.
2606
2607	Default: 16
2608
2609disable_ipv6 - BOOLEAN
2610	Disable IPv6 operation.  If accept_dad is set to 2, this value
2611	will be dynamically set to TRUE if DAD fails for the link-local
2612	address.
2613
2614	Default: FALSE (enable IPv6 operation)
2615
2616	When this value is changed from 1 to 0 (IPv6 is being enabled),
2617	it will dynamically create a link-local address on the given
2618	interface and start Duplicate Address Detection, if necessary.
2619
2620	When this value is changed from 0 to 1 (IPv6 is being disabled),
2621	it will dynamically delete all addresses and routes on the given
2622	interface. From now on it will not possible to add addresses/routes
2623	to the selected interface.
2624
2625accept_dad - INTEGER
2626	Whether to accept DAD (Duplicate Address Detection).
2627
2628	 == ==============================================================
2629	  0  Disable DAD
2630	  1  Enable DAD (default)
2631	  2  Enable DAD, and disable IPv6 operation if MAC-based duplicate
2632	     link-local address has been found.
2633	 == ==============================================================
2634
2635	DAD operation and mode on a given interface will be selected according
2636	to the maximum value of conf/{all,interface}/accept_dad.
2637
2638force_tllao - BOOLEAN
2639	Enable sending the target link-layer address option even when
2640	responding to a unicast neighbor solicitation.
2641
2642	Default: FALSE
2643
2644	Quoting from RFC 2461, section 4.4, Target link-layer address:
2645
2646	"The option MUST be included for multicast solicitations in order to
2647	avoid infinite Neighbor Solicitation "recursion" when the peer node
2648	does not have a cache entry to return a Neighbor Advertisements
2649	message.  When responding to unicast solicitations, the option can be
2650	omitted since the sender of the solicitation has the correct link-
2651	layer address; otherwise it would not have be able to send the unicast
2652	solicitation in the first place. However, including the link-layer
2653	address in this case adds little overhead and eliminates a potential
2654	race condition where the sender deletes the cached link-layer address
2655	prior to receiving a response to a previous solicitation."
2656
2657ndisc_notify - BOOLEAN
2658	Define mode for notification of address and device changes.
2659
2660	* 0 - (default): do nothing
2661	* 1 - Generate unsolicited neighbour advertisements when device is brought
2662	  up or hardware address changes.
2663
2664ndisc_tclass - INTEGER
2665	The IPv6 Traffic Class to use by default when sending IPv6 Neighbor
2666	Discovery (Router Solicitation, Router Advertisement, Neighbor
2667	Solicitation, Neighbor Advertisement, Redirect) messages.
2668	These 8 bits can be interpreted as 6 high order bits holding the DSCP
2669	value and 2 low order bits representing ECN (which you probably want
2670	to leave cleared).
2671
2672	* 0 - (default)
2673
2674ndisc_evict_nocarrier - BOOLEAN
2675	Clears the neighbor discovery table on NOCARRIER events. This option is
2676	important for wireless devices where the neighbor discovery cache should
2677	not be cleared when roaming between access points on the same network.
2678	In most cases this should remain as the default (1).
2679
2680	- 1 - (default): Clear neighbor discover cache on NOCARRIER events.
2681	- 0 - Do not clear neighbor discovery cache on NOCARRIER events.
2682
2683mldv1_unsolicited_report_interval - INTEGER
2684	The interval in milliseconds in which the next unsolicited
2685	MLDv1 report retransmit will take place.
2686
2687	Default: 10000 (10 seconds)
2688
2689mldv2_unsolicited_report_interval - INTEGER
2690	The interval in milliseconds in which the next unsolicited
2691	MLDv2 report retransmit will take place.
2692
2693	Default: 1000 (1 second)
2694
2695force_mld_version - INTEGER
2696	* 0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
2697	* 1 - Enforce to use MLD version 1
2698	* 2 - Enforce to use MLD version 2
2699
2700suppress_frag_ndisc - INTEGER
2701	Control RFC 6980 (Security Implications of IPv6 Fragmentation
2702	with IPv6 Neighbor Discovery) behavior:
2703
2704	* 1 - (default) discard fragmented neighbor discovery packets
2705	* 0 - allow fragmented neighbor discovery packets
2706
2707optimistic_dad - BOOLEAN
2708	Whether to perform Optimistic Duplicate Address Detection (RFC 4429).
2709
2710	* 0: disabled (default)
2711	* 1: enabled
2712
2713	Optimistic Duplicate Address Detection for the interface will be enabled
2714	if at least one of conf/{all,interface}/optimistic_dad is set to 1,
2715	it will be disabled otherwise.
2716
2717use_optimistic - BOOLEAN
2718	If enabled, do not classify optimistic addresses as deprecated during
2719	source address selection.  Preferred addresses will still be chosen
2720	before optimistic addresses, subject to other ranking in the source
2721	address selection algorithm.
2722
2723	* 0: disabled (default)
2724	* 1: enabled
2725
2726	This will be enabled if at least one of
2727	conf/{all,interface}/use_optimistic is set to 1, disabled otherwise.
2728
2729stable_secret - IPv6 address
2730	This IPv6 address will be used as a secret to generate IPv6
2731	addresses for link-local addresses and autoconfigured
2732	ones. All addresses generated after setting this secret will
2733	be stable privacy ones by default. This can be changed via the
2734	addrgenmode ip-link. conf/default/stable_secret is used as the
2735	secret for the namespace, the interface specific ones can
2736	overwrite that. Writes to conf/all/stable_secret are refused.
2737
2738	It is recommended to generate this secret during installation
2739	of a system and keep it stable after that.
2740
2741	By default the stable secret is unset.
2742
2743addr_gen_mode - INTEGER
2744	Defines how link-local and autoconf addresses are generated.
2745
2746	=  =================================================================
2747	0  generate address based on EUI64 (default)
2748	1  do no generate a link-local address, use EUI64 for addresses
2749	   generated from autoconf
2750	2  generate stable privacy addresses, using the secret from
2751	   stable_secret (RFC7217)
2752	3  generate stable privacy addresses, using a random secret if unset
2753	=  =================================================================
2754
2755drop_unicast_in_l2_multicast - BOOLEAN
2756	Drop any unicast IPv6 packets that are received in link-layer
2757	multicast (or broadcast) frames.
2758
2759	By default this is turned off.
2760
2761drop_unsolicited_na - BOOLEAN
2762	Drop all unsolicited neighbor advertisements, for example if there's
2763	a known good NA proxy on the network and such frames need not be used
2764	(or in the case of 802.11, must not be used to prevent attacks.)
2765
2766	By default this is turned off.
2767
2768accept_untracked_na - INTEGER
2769	Define behavior for accepting neighbor advertisements from devices that
2770	are absent in the neighbor cache:
2771
2772	- 0 - (default) Do not accept unsolicited and untracked neighbor
2773	  advertisements.
2774
2775	- 1 - Add a new neighbor cache entry in STALE state for routers on
2776	  receiving a neighbor advertisement (either solicited or unsolicited)
2777	  with target link-layer address option specified if no neighbor entry
2778	  is already present for the advertised IPv6 address. Without this knob,
2779	  NAs received for untracked addresses (absent in neighbor cache) are
2780	  silently ignored.
2781
2782	  This is as per router-side behavior documented in RFC9131.
2783
2784	  This has lower precedence than drop_unsolicited_na.
2785
2786	  This will optimize the return path for the initial off-link
2787	  communication that is initiated by a directly connected host, by
2788	  ensuring that the first-hop router which turns on this setting doesn't
2789	  have to buffer the initial return packets to do neighbor-solicitation.
2790	  The prerequisite is that the host is configured to send unsolicited
2791	  neighbor advertisements on interface bringup. This setting should be
2792	  used in conjunction with the ndisc_notify setting on the host to
2793	  satisfy this prerequisite.
2794
2795	- 2 - Extend option (1) to add a new neighbor cache entry only if the
2796	  source IP address is in the same subnet as an address configured on
2797	  the interface that received the neighbor advertisement.
2798
2799enhanced_dad - BOOLEAN
2800	Include a nonce option in the IPv6 neighbor solicitation messages used for
2801	duplicate address detection per RFC7527. A received DAD NS will only signal
2802	a duplicate address if the nonce is different. This avoids any false
2803	detection of duplicates due to loopback of the NS messages that we send.
2804	The nonce option will be sent on an interface unless both of
2805	conf/{all,interface}/enhanced_dad are set to FALSE.
2806
2807	Default: TRUE
2808
2809``icmp/*``:
2810===========
2811
2812ratelimit - INTEGER
2813	Limit the maximal rates for sending ICMPv6 messages.
2814
2815	0 to disable any limiting,
2816	otherwise the minimal space between responses in milliseconds.
2817
2818	Default: 1000
2819
2820ratemask - list of comma separated ranges
2821	For ICMPv6 message types matching the ranges in the ratemask, limit
2822	the sending of the message according to ratelimit parameter.
2823
2824	The format used for both input and output is a comma separated
2825	list of ranges (e.g. "0-127,129" for ICMPv6 message type 0 to 127 and
2826	129). Writing to the file will clear all previous ranges of ICMPv6
2827	message types and update the current list with the input.
2828
2829	Refer to: https://www.iana.org/assignments/icmpv6-parameters/icmpv6-parameters.xhtml
2830	for numerical values of ICMPv6 message types, e.g. echo request is 128
2831	and echo reply is 129.
2832
2833	Default: 0-1,3-127 (rate limit ICMPv6 errors except Packet Too Big)
2834
2835echo_ignore_all - BOOLEAN
2836	If set non-zero, then the kernel will ignore all ICMP ECHO
2837	requests sent to it over the IPv6 protocol.
2838
2839	Default: 0
2840
2841echo_ignore_multicast - BOOLEAN
2842	If set non-zero, then the kernel will ignore all ICMP ECHO
2843	requests sent to it over the IPv6 protocol via multicast.
2844
2845	Default: 0
2846
2847echo_ignore_anycast - BOOLEAN
2848	If set non-zero, then the kernel will ignore all ICMP ECHO
2849	requests sent to it over the IPv6 protocol destined to anycast address.
2850
2851	Default: 0
2852
2853error_anycast_as_unicast - BOOLEAN
2854	If set to 1, then the kernel will respond with ICMP Errors
2855	resulting from requests sent to it over the IPv6 protocol destined
2856	to anycast address essentially treating anycast as unicast.
2857
2858	Default: 0
2859
2860xfrm6_gc_thresh - INTEGER
2861	(Obsolete since linux-4.14)
2862	The threshold at which we will start garbage collecting for IPv6
2863	destination cache entries.  At twice this value the system will
2864	refuse new allocations.
2865
2866
2867IPv6 Update by:
2868Pekka Savola <pekkas@netcore.fi>
2869YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
2870
2871
2872/proc/sys/net/bridge/* Variables:
2873=================================
2874
2875bridge-nf-call-arptables - BOOLEAN
2876	- 1 : pass bridged ARP traffic to arptables' FORWARD chain.
2877	- 0 : disable this.
2878
2879	Default: 1
2880
2881bridge-nf-call-iptables - BOOLEAN
2882	- 1 : pass bridged IPv4 traffic to iptables' chains.
2883	- 0 : disable this.
2884
2885	Default: 1
2886
2887bridge-nf-call-ip6tables - BOOLEAN
2888	- 1 : pass bridged IPv6 traffic to ip6tables' chains.
2889	- 0 : disable this.
2890
2891	Default: 1
2892
2893bridge-nf-filter-vlan-tagged - BOOLEAN
2894	- 1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
2895	- 0 : disable this.
2896
2897	Default: 0
2898
2899bridge-nf-filter-pppoe-tagged - BOOLEAN
2900	- 1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
2901	- 0 : disable this.
2902
2903	Default: 0
2904
2905bridge-nf-pass-vlan-input-dev - BOOLEAN
2906	- 1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
2907	  interface on the bridge and set the netfilter input device to the
2908	  vlan. This allows use of e.g. "iptables -i br0.1" and makes the
2909	  REDIRECT target work with vlan-on-top-of-bridge interfaces.  When no
2910	  matching vlan interface is found, or this switch is off, the input
2911	  device is set to the bridge interface.
2912
2913	- 0: disable bridge netfilter vlan interface lookup.
2914
2915	Default: 0
2916
2917``proc/sys/net/sctp/*`` Variables:
2918==================================
2919
2920addip_enable - BOOLEAN
2921	Enable or disable extension of  Dynamic Address Reconfiguration
2922	(ADD-IP) functionality specified in RFC5061.  This extension provides
2923	the ability to dynamically add and remove new addresses for the SCTP
2924	associations.
2925
2926	1: Enable extension.
2927
2928	0: Disable extension.
2929
2930	Default: 0
2931
2932pf_enable - INTEGER
2933	Enable or disable pf (pf is short for potentially failed) state. A value
2934	of pf_retrans > path_max_retrans also disables pf state. That is, one of
2935	both pf_enable and pf_retrans > path_max_retrans can disable pf state.
2936	Since pf_retrans and path_max_retrans can be changed by userspace
2937	application, sometimes user expects to disable pf state by the value of
2938	pf_retrans > path_max_retrans, but occasionally the value of pf_retrans
2939	or path_max_retrans is changed by the user application, this pf state is
2940	enabled. As such, it is necessary to add this to dynamically enable
2941	and disable pf state. See:
2942	https://datatracker.ietf.org/doc/draft-ietf-tsvwg-sctp-failover for
2943	details.
2944
2945	1: Enable pf.
2946
2947	0: Disable pf.
2948
2949	Default: 1
2950
2951pf_expose - INTEGER
2952	Unset or enable/disable pf (pf is short for potentially failed) state
2953	exposure.  Applications can control the exposure of the PF path state
2954	in the SCTP_PEER_ADDR_CHANGE event and the SCTP_GET_PEER_ADDR_INFO
2955	sockopt.   When it's unset, no SCTP_PEER_ADDR_CHANGE event with
2956	SCTP_ADDR_PF state will be sent and a SCTP_PF-state transport info
2957	can be got via SCTP_GET_PEER_ADDR_INFO sockopt;  When it's enabled,
2958	a SCTP_PEER_ADDR_CHANGE event will be sent for a transport becoming
2959	SCTP_PF state and a SCTP_PF-state transport info can be got via
2960	SCTP_GET_PEER_ADDR_INFO sockopt;  When it's disabled, no
2961	SCTP_PEER_ADDR_CHANGE event will be sent and it returns -EACCES when
2962	trying to get a SCTP_PF-state transport info via SCTP_GET_PEER_ADDR_INFO
2963	sockopt.
2964
2965	0: Unset pf state exposure, Compatible with old applications.
2966
2967	1: Disable pf state exposure.
2968
2969	2: Enable pf state exposure.
2970
2971	Default: 0
2972
2973addip_noauth_enable - BOOLEAN
2974	Dynamic Address Reconfiguration (ADD-IP) requires the use of
2975	authentication to protect the operations of adding or removing new
2976	addresses.  This requirement is mandated so that unauthorized hosts
2977	would not be able to hijack associations.  However, older
2978	implementations may not have implemented this requirement while
2979	allowing the ADD-IP extension.  For reasons of interoperability,
2980	we provide this variable to control the enforcement of the
2981	authentication requirement.
2982
2983	== ===============================================================
2984	1  Allow ADD-IP extension to be used without authentication.  This
2985	   should only be set in a closed environment for interoperability
2986	   with older implementations.
2987
2988	0  Enforce the authentication requirement
2989	== ===============================================================
2990
2991	Default: 0
2992
2993auth_enable - BOOLEAN
2994	Enable or disable Authenticated Chunks extension.  This extension
2995	provides the ability to send and receive authenticated chunks and is
2996	required for secure operation of Dynamic Address Reconfiguration
2997	(ADD-IP) extension.
2998
2999	- 1: Enable this extension.
3000	- 0: Disable this extension.
3001
3002	Default: 0
3003
3004prsctp_enable - BOOLEAN
3005	Enable or disable the Partial Reliability extension (RFC3758) which
3006	is used to notify peers that a given DATA should no longer be expected.
3007
3008	- 1: Enable extension
3009	- 0: Disable
3010
3011	Default: 1
3012
3013max_burst - INTEGER
3014	The limit of the number of new packets that can be initially sent.  It
3015	controls how bursty the generated traffic can be.
3016
3017	Default: 4
3018
3019association_max_retrans - INTEGER
3020	Set the maximum number for retransmissions that an association can
3021	attempt deciding that the remote end is unreachable.  If this value
3022	is exceeded, the association is terminated.
3023
3024	Default: 10
3025
3026max_init_retransmits - INTEGER
3027	The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
3028	that an association will attempt before declaring the destination
3029	unreachable and terminating.
3030
3031	Default: 8
3032
3033path_max_retrans - INTEGER
3034	The maximum number of retransmissions that will be attempted on a given
3035	path.  Once this threshold is exceeded, the path is considered
3036	unreachable, and new traffic will use a different path when the
3037	association is multihomed.
3038
3039	Default: 5
3040
3041pf_retrans - INTEGER
3042	The number of retransmissions that will be attempted on a given path
3043	before traffic is redirected to an alternate transport (should one
3044	exist).  Note this is distinct from path_max_retrans, as a path that
3045	passes the pf_retrans threshold can still be used.  Its only
3046	deprioritized when a transmission path is selected by the stack.  This
3047	setting is primarily used to enable fast failover mechanisms without
3048	having to reduce path_max_retrans to a very low value.  See:
3049	http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3050	for details.  Note also that a value of pf_retrans > path_max_retrans
3051	disables this feature. Since both pf_retrans and path_max_retrans can
3052	be changed by userspace application, a variable pf_enable is used to
3053	disable pf state.
3054
3055	Default: 0
3056
3057ps_retrans - INTEGER
3058	Primary.Switchover.Max.Retrans (PSMR), it's a tunable parameter coming
3059	from section-5 "Primary Path Switchover" in rfc7829.  The primary path
3060	will be changed to another active path when the path error counter on
3061	the old primary path exceeds PSMR, so that "the SCTP sender is allowed
3062	to continue data transmission on a new working path even when the old
3063	primary destination address becomes active again".   Note this feature
3064	is disabled by initializing 'ps_retrans' per netns as 0xffff by default,
3065	and its value can't be less than 'pf_retrans' when changing by sysctl.
3066
3067	Default: 0xffff
3068
3069rto_initial - INTEGER
3070	The initial round trip timeout value in milliseconds that will be used
3071	in calculating round trip times.  This is the initial time interval
3072	for retransmissions.
3073
3074	Default: 3000
3075
3076rto_max - INTEGER
3077	The maximum value (in milliseconds) of the round trip timeout.  This
3078	is the largest time interval that can elapse between retransmissions.
3079
3080	Default: 60000
3081
3082rto_min - INTEGER
3083	The minimum value (in milliseconds) of the round trip timeout.  This
3084	is the smallest time interval the can elapse between retransmissions.
3085
3086	Default: 1000
3087
3088hb_interval - INTEGER
3089	The interval (in milliseconds) between HEARTBEAT chunks.  These chunks
3090	are sent at the specified interval on idle paths to probe the state of
3091	a given path between 2 associations.
3092
3093	Default: 30000
3094
3095sack_timeout - INTEGER
3096	The amount of time (in milliseconds) that the implementation will wait
3097	to send a SACK.
3098
3099	Default: 200
3100
3101valid_cookie_life - INTEGER
3102	The default lifetime of the SCTP cookie (in milliseconds).  The cookie
3103	is used during association establishment.
3104
3105	Default: 60000
3106
3107cookie_preserve_enable - BOOLEAN
3108	Enable or disable the ability to extend the lifetime of the SCTP cookie
3109	that is used during the establishment phase of SCTP association
3110
3111	- 1: Enable cookie lifetime extension.
3112	- 0: Disable
3113
3114	Default: 1
3115
3116cookie_hmac_alg - STRING
3117	Select the hmac algorithm used when generating the cookie value sent by
3118	a listening sctp socket to a connecting client in the INIT-ACK chunk.
3119	Valid values are:
3120
3121	* md5
3122	* sha1
3123	* none
3124
3125	Ability to assign md5 or sha1 as the selected alg is predicated on the
3126	configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
3127	CONFIG_CRYPTO_SHA1).
3128
3129	Default: Dependent on configuration.  MD5 if available, else SHA1 if
3130	available, else none.
3131
3132rcvbuf_policy - INTEGER
3133	Determines if the receive buffer is attributed to the socket or to
3134	association.   SCTP supports the capability to create multiple
3135	associations on a single socket.  When using this capability, it is
3136	possible that a single stalled association that's buffering a lot
3137	of data may block other associations from delivering their data by
3138	consuming all of the receive buffer space.  To work around this,
3139	the rcvbuf_policy could be set to attribute the receiver buffer space
3140	to each association instead of the socket.  This prevents the described
3141	blocking.
3142
3143	- 1: rcvbuf space is per association
3144	- 0: rcvbuf space is per socket
3145
3146	Default: 0
3147
3148sndbuf_policy - INTEGER
3149	Similar to rcvbuf_policy above, this applies to send buffer space.
3150
3151	- 1: Send buffer is tracked per association
3152	- 0: Send buffer is tracked per socket.
3153
3154	Default: 0
3155
3156sctp_mem - vector of 3 INTEGERs: min, pressure, max
3157	Number of pages allowed for queueing by all SCTP sockets.
3158
3159	min: Below this number of pages SCTP is not bothered about its
3160	memory appetite. When amount of memory allocated by SCTP exceeds
3161	this number, SCTP starts to moderate memory usage.
3162
3163	pressure: This value was introduced to follow format of tcp_mem.
3164
3165	max: Number of pages allowed for queueing by all SCTP sockets.
3166
3167	Default is calculated at boot time from amount of available memory.
3168
3169sctp_rmem - vector of 3 INTEGERs: min, default, max
3170	Only the first value ("min") is used, "default" and "max" are
3171	ignored.
3172
3173	min: Minimal size of receive buffer used by SCTP socket.
3174	It is guaranteed to each SCTP socket (but not association) even
3175	under moderate memory pressure.
3176
3177	Default: 4K
3178
3179sctp_wmem  - vector of 3 INTEGERs: min, default, max
3180	Only the first value ("min") is used, "default" and "max" are
3181	ignored.
3182
3183	min: Minimum size of send buffer that can be used by SCTP sockets.
3184	It is guaranteed to each SCTP socket (but not association) even
3185	under moderate memory pressure.
3186
3187	Default: 4K
3188
3189addr_scope_policy - INTEGER
3190	Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
3191
3192	- 0   - Disable IPv4 address scoping
3193	- 1   - Enable IPv4 address scoping
3194	- 2   - Follow draft but allow IPv4 private addresses
3195	- 3   - Follow draft but allow IPv4 link local addresses
3196
3197	Default: 1
3198
3199udp_port - INTEGER
3200	The listening port for the local UDP tunneling sock. Normally it's
3201	using the IANA-assigned UDP port number 9899 (sctp-tunneling).
3202
3203	This UDP sock is used for processing the incoming UDP-encapsulated
3204	SCTP packets (from RFC6951), and shared by all applications in the
3205	same net namespace. This UDP sock will be closed when the value is
3206	set to 0.
3207
3208	The value will also be used to set the src port of the UDP header
3209	for the outgoing UDP-encapsulated SCTP packets. For the dest port,
3210	please refer to 'encap_port' below.
3211
3212	Default: 0
3213
3214encap_port - INTEGER
3215	The default remote UDP encapsulation port.
3216
3217	This value is used to set the dest port of the UDP header for the
3218	outgoing UDP-encapsulated SCTP packets by default. Users can also
3219	change the value for each sock/asoc/transport by using setsockopt.
3220	For further information, please refer to RFC6951.
3221
3222	Note that when connecting to a remote server, the client should set
3223	this to the port that the UDP tunneling sock on the peer server is
3224	listening to and the local UDP tunneling sock on the client also
3225	must be started. On the server, it would get the encap_port from
3226	the incoming packet's source port.
3227
3228	Default: 0
3229
3230plpmtud_probe_interval - INTEGER
3231        The time interval (in milliseconds) for the PLPMTUD probe timer,
3232        which is configured to expire after this period to receive an
3233        acknowledgment to a probe packet. This is also the time interval
3234        between the probes for the current pmtu when the probe search
3235        is done.
3236
3237        PLPMTUD will be disabled when 0 is set, and other values for it
3238        must be >= 5000.
3239
3240	Default: 0
3241
3242reconf_enable - BOOLEAN
3243        Enable or disable extension of Stream Reconfiguration functionality
3244        specified in RFC6525. This extension provides the ability to "reset"
3245        a stream, and it includes the Parameters of "Outgoing/Incoming SSN
3246        Reset", "SSN/TSN Reset" and "Add Outgoing/Incoming Streams".
3247
3248	- 1: Enable extension.
3249	- 0: Disable extension.
3250
3251	Default: 0
3252
3253intl_enable - BOOLEAN
3254        Enable or disable extension of User Message Interleaving functionality
3255        specified in RFC8260. This extension allows the interleaving of user
3256        messages sent on different streams. With this feature enabled, I-DATA
3257        chunk will replace DATA chunk to carry user messages if also supported
3258        by the peer. Note that to use this feature, one needs to set this option
3259        to 1 and also needs to set socket options SCTP_FRAGMENT_INTERLEAVE to 2
3260        and SCTP_INTERLEAVING_SUPPORTED to 1.
3261
3262	- 1: Enable extension.
3263	- 0: Disable extension.
3264
3265	Default: 0
3266
3267ecn_enable - BOOLEAN
3268        Control use of Explicit Congestion Notification (ECN) by SCTP.
3269        Like in TCP, ECN is used only when both ends of the SCTP connection
3270        indicate support for it. This feature is useful in avoiding losses
3271        due to congestion by allowing supporting routers to signal congestion
3272        before having to drop packets.
3273
3274        1: Enable ecn.
3275        0: Disable ecn.
3276
3277        Default: 1
3278
3279l3mdev_accept - BOOLEAN
3280	Enabling this option allows a "global" bound socket to work
3281	across L3 master domains (e.g., VRFs) with packets capable of
3282	being received regardless of the L3 domain in which they
3283	originated. Only valid when the kernel was compiled with
3284	CONFIG_NET_L3_MASTER_DEV.
3285
3286	Default: 1 (enabled)
3287
3288
3289``/proc/sys/net/core/*``
3290========================
3291
3292	Please see: Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
3293
3294
3295``/proc/sys/net/unix/*``
3296========================
3297
3298max_dgram_qlen - INTEGER
3299	The maximum length of dgram socket receive queue
3300
3301	Default: 10
3302