linux/Documentation/networking/ip-sysctl.txt
<<
>>
Prefs
   1/proc/sys/net/ipv4/* Variables:
   2
   3ip_forward - BOOLEAN
   4        0 - disabled (default)
   5        not 0 - enabled
   6
   7        Forward Packets between interfaces.
   8
   9        This variable is special, its change resets all configuration
  10        parameters to their default state (RFC1122 for hosts, RFC1812
  11        for routers)
  12
  13ip_default_ttl - INTEGER
  14        Default value of TTL field (Time To Live) for outgoing (but not
  15        forwarded) IP packets. Should be between 1 and 255 inclusive.
  16        Default: 64 (as recommended by RFC1700)
  17
  18ip_no_pmtu_disc - BOOLEAN
  19        Disable Path MTU Discovery.
  20        default FALSE
  21
  22min_pmtu - INTEGER
  23        default 552 - minimum discovered Path MTU
  24
  25route/max_size - INTEGER
  26        Maximum number of routes allowed in the kernel.  Increase
  27        this when using large numbers of interfaces and/or routes.
  28
  29neigh/default/gc_thresh3 - INTEGER
  30        Maximum number of neighbor entries allowed.  Increase this
  31        when using large numbers of interfaces and when communicating
  32        with large numbers of directly-connected peers.
  33
  34neigh/default/unres_qlen_bytes - INTEGER
  35        The maximum number of bytes which may be used by packets
  36        queued for each unresolved address by other network layers.
  37        (added in linux 3.3)
  38
  39neigh/default/unres_qlen - INTEGER
  40        The maximum number of packets which may be queued for each
  41        unresolved address by other network layers.
  42        (deprecated in linux 3.3) : use unres_qlen_bytes instead.
  43
  44mtu_expires - INTEGER
  45        Time, in seconds, that cached PMTU information is kept.
  46
  47min_adv_mss - INTEGER
  48        The advertised MSS depends on the first hop route MTU, but will
  49        never be lower than this setting.
  50
  51IP Fragmentation:
  52
  53ipfrag_high_thresh - INTEGER
  54        Maximum memory used to reassemble IP fragments. When
  55        ipfrag_high_thresh bytes of memory is allocated for this purpose,
  56        the fragment handler will toss packets until ipfrag_low_thresh
  57        is reached.
  58
  59ipfrag_low_thresh - INTEGER
  60        See ipfrag_high_thresh
  61
  62ipfrag_time - INTEGER
  63        Time in seconds to keep an IP fragment in memory.
  64
  65ipfrag_secret_interval - INTEGER
  66        Regeneration interval (in seconds) of the hash secret (or lifetime
  67        for the hash secret) for IP fragments.
  68        Default: 600
  69
  70ipfrag_max_dist - INTEGER
  71        ipfrag_max_dist is a non-negative integer value which defines the
  72        maximum "disorder" which is allowed among fragments which share a
  73        common IP source address. Note that reordering of packets is
  74        not unusual, but if a large number of fragments arrive from a source
  75        IP address while a particular fragment queue remains incomplete, it
  76        probably indicates that one or more fragments belonging to that queue
  77        have been lost. When ipfrag_max_dist is positive, an additional check
  78        is done on fragments before they are added to a reassembly queue - if
  79        ipfrag_max_dist (or more) fragments have arrived from a particular IP
  80        address between additions to any IP fragment queue using that source
  81        address, it's presumed that one or more fragments in the queue are
  82        lost. The existing fragment queue will be dropped, and a new one
  83        started. An ipfrag_max_dist value of zero disables this check.
  84
  85        Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
  86        result in unnecessarily dropping fragment queues when normal
  87        reordering of packets occurs, which could lead to poor application
  88        performance. Using a very large value, e.g. 50000, increases the
  89        likelihood of incorrectly reassembling IP fragments that originate
  90        from different IP datagrams, which could result in data corruption.
  91        Default: 64
  92
  93INET peer storage:
  94
  95inet_peer_threshold - INTEGER
  96        The approximate size of the storage.  Starting from this threshold
  97        entries will be thrown aggressively.  This threshold also determines
  98        entries' time-to-live and time intervals between garbage collection
  99        passes.  More entries, less time-to-live, less GC interval.
 100
 101inet_peer_minttl - INTEGER
 102        Minimum time-to-live of entries.  Should be enough to cover fragment
 103        time-to-live on the reassembling side.  This minimum time-to-live  is
 104        guaranteed if the pool size is less than inet_peer_threshold.
 105        Measured in seconds.
 106
 107inet_peer_maxttl - INTEGER
 108        Maximum time-to-live of entries.  Unused entries will expire after
 109        this period of time if there is no memory pressure on the pool (i.e.
 110        when the number of entries in the pool is very small).
 111        Measured in seconds.
 112
 113TCP variables:
 114
 115somaxconn - INTEGER
 116        Limit of socket listen() backlog, known in userspace as SOMAXCONN.
 117        Defaults to 128.  See also tcp_max_syn_backlog for additional tuning
 118        for TCP sockets.
 119
 120tcp_abc - INTEGER
 121        Controls Appropriate Byte Count (ABC) defined in RFC3465.
 122        ABC is a way of increasing congestion window (cwnd) more slowly
 123        in response to partial acknowledgments.
 124        Possible values are:
 125                0 increase cwnd once per acknowledgment (no ABC)
 126                1 increase cwnd once per acknowledgment of full sized segment
 127                2 allow increase cwnd by two if acknowledgment is
 128                  of two segments to compensate for delayed acknowledgments.
 129        Default: 0 (off)
 130
 131tcp_abort_on_overflow - BOOLEAN
 132        If listening service is too slow to accept new connections,
 133        reset them. Default state is FALSE. It means that if overflow
 134        occurred due to a burst, connection will recover. Enable this
 135        option _only_ if you are really sure that listening daemon
 136        cannot be tuned to accept connections faster. Enabling this
 137        option can harm clients of your server.
 138
 139tcp_adv_win_scale - INTEGER
 140        Count buffering overhead as bytes/2^tcp_adv_win_scale
 141        (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
 142        if it is <= 0.
 143        Possible values are [-31, 31], inclusive.
 144        Default: 1
 145
 146tcp_allowed_congestion_control - STRING
 147        Show/set the congestion control choices available to non-privileged
 148        processes. The list is a subset of those listed in
 149        tcp_available_congestion_control.
 150        Default is "reno" and the default setting (tcp_congestion_control).
 151
 152tcp_app_win - INTEGER
 153        Reserve max(window/2^tcp_app_win, mss) of window for application
 154        buffer. Value 0 is special, it means that nothing is reserved.
 155        Default: 31
 156
 157tcp_available_congestion_control - STRING
 158        Shows the available congestion control choices that are registered.
 159        More congestion control algorithms may be available as modules,
 160        but not loaded.
 161
 162tcp_base_mss - INTEGER
 163        The initial value of search_low to be used by the packetization layer
 164        Path MTU discovery (MTU probing).  If MTU probing is enabled,
 165        this is the initial MSS used by the connection.
 166
 167tcp_congestion_control - STRING
 168        Set the congestion control algorithm to be used for new
 169        connections. The algorithm "reno" is always available, but
 170        additional choices may be available based on kernel configuration.
 171        Default is set as part of kernel configuration.
 172        For passive connections, the listener congestion control choice
 173        is inherited.
 174        [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
 175
 176tcp_cookie_size - INTEGER
 177        Default size of TCP Cookie Transactions (TCPCT) option, that may be
 178        overridden on a per socket basis by the TCPCT socket option.
 179        Values greater than the maximum (16) are interpreted as the maximum.
 180        Values greater than zero and less than the minimum (8) are interpreted
 181        as the minimum.  Odd values are interpreted as the next even value.
 182        Default: 0 (off).
 183
 184tcp_dsack - BOOLEAN
 185        Allows TCP to send "duplicate" SACKs.
 186
 187tcp_early_retrans - INTEGER
 188        Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
 189        for triggering fast retransmit when the amount of outstanding data is
 190        small and when no previously unsent data can be transmitted (such
 191        that limited transmit could be used).
 192        Possible values:
 193                0 disables ER
 194                1 enables ER
 195                2 enables ER but delays fast recovery and fast retransmit
 196                  by a fourth of RTT. This mitigates connection falsely
 197                  recovers when network has a small degree of reordering
 198                  (less than 3 packets).
 199        Default: 2
 200
 201tcp_ecn - INTEGER
 202        Enable Explicit Congestion Notification (ECN) in TCP. ECN is only
 203        used when both ends of the TCP flow support it. It is useful to
 204        avoid losses due to congestion (when the bottleneck router supports
 205        ECN).
 206        Possible values are:
 207                0 disable ECN
 208                1 ECN enabled
 209                2 Only server-side ECN enabled. If the other end does
 210                  not support ECN, behavior is like with ECN disabled.
 211        Default: 2
 212
 213tcp_fack - BOOLEAN
 214        Enable FACK congestion avoidance and fast retransmission.
 215        The value is not used, if tcp_sack is not enabled.
 216
 217tcp_fin_timeout - INTEGER
 218        Time to hold socket in state FIN-WAIT-2, if it was closed
 219        by our side. Peer can be broken and never close its side,
 220        or even died unexpectedly. Default value is 60sec.
 221        Usual value used in 2.2 was 180 seconds, you may restore
 222        it, but remember that if your machine is even underloaded WEB server,
 223        you risk to overflow memory with kilotons of dead sockets,
 224        FIN-WAIT-2 sockets are less dangerous than FIN-WAIT-1,
 225        because they eat maximum 1.5K of memory, but they tend
 226        to live longer. Cf. tcp_max_orphans.
 227
 228tcp_frto - INTEGER
 229        Enables Forward RTO-Recovery (F-RTO) defined in RFC4138.
 230        F-RTO is an enhanced recovery algorithm for TCP retransmission
 231        timeouts.  It is particularly beneficial in wireless environments
 232        where packet loss is typically due to random radio interference
 233        rather than intermediate router congestion.  F-RTO is sender-side
 234        only modification. Therefore it does not require any support from
 235        the peer.
 236
 237        If set to 1, basic version is enabled.  2 enables SACK enhanced
 238        F-RTO if flow uses SACK.  The basic version can be used also when
 239        SACK is in use though scenario(s) with it exists where F-RTO
 240        interacts badly with the packet counting of the SACK enabled TCP
 241        flow.
 242
 243tcp_frto_response - INTEGER
 244        When F-RTO has detected that a TCP retransmission timeout was
 245        spurious (i.e, the timeout would have been avoided had TCP set a
 246        longer retransmission timeout), TCP has several options what to do
 247        next. Possible values are:
 248                0 Rate halving based; a smooth and conservative response,
 249                  results in halved cwnd and ssthresh after one RTT
 250                1 Very conservative response; not recommended because even
 251                  though being valid, it interacts poorly with the rest of
 252                  Linux TCP, halves cwnd and ssthresh immediately
 253                2 Aggressive response; undoes congestion control measures
 254                  that are now known to be unnecessary (ignoring the
 255                  possibility of a lost retransmission that would require
 256                  TCP to be more cautious), cwnd and ssthresh are restored
 257                  to the values prior timeout
 258        Default: 0 (rate halving based)
 259
 260tcp_keepalive_time - INTEGER
 261        How often TCP sends out keepalive messages when keepalive is enabled.
 262        Default: 2hours.
 263
 264tcp_keepalive_probes - INTEGER
 265        How many keepalive probes TCP sends out, until it decides that the
 266        connection is broken. Default value: 9.
 267
 268tcp_keepalive_intvl - INTEGER
 269        How frequently the probes are send out. Multiplied by
 270        tcp_keepalive_probes it is time to kill not responding connection,
 271        after probes started. Default value: 75sec i.e. connection
 272        will be aborted after ~11 minutes of retries.
 273
 274tcp_low_latency - BOOLEAN
 275        If set, the TCP stack makes decisions that prefer lower
 276        latency as opposed to higher throughput.  By default, this
 277        option is not set meaning that higher throughput is preferred.
 278        An example of an application where this default should be
 279        changed would be a Beowulf compute cluster.
 280        Default: 0
 281
 282tcp_max_orphans - INTEGER
 283        Maximal number of TCP sockets not attached to any user file handle,
 284        held by system. If this number is exceeded orphaned connections are
 285        reset immediately and warning is printed. This limit exists
 286        only to prevent simple DoS attacks, you _must_ not rely on this
 287        or lower the limit artificially, but rather increase it
 288        (probably, after increasing installed memory),
 289        if network conditions require more than default value,
 290        and tune network services to linger and kill such states
 291        more aggressively. Let me to remind again: each orphan eats
 292        up to ~64K of unswappable memory.
 293
 294tcp_max_ssthresh - INTEGER
 295        Limited Slow-Start for TCP with large congestion windows (cwnd) defined in
 296        RFC3742. Limited slow-start is a mechanism to limit growth of the cwnd
 297        on the region where cwnd is larger than tcp_max_ssthresh. TCP increases cwnd
 298        by at most tcp_max_ssthresh segments, and by at least tcp_max_ssthresh/2
 299        segments per RTT when the cwnd is above tcp_max_ssthresh.
 300        If TCP connection increased cwnd to thousands (or tens of thousands) segments,
 301        and thousands of packets were being dropped during slow-start, you can set
 302        tcp_max_ssthresh to improve performance for new TCP connection.
 303        Default: 0 (off)
 304
 305tcp_max_syn_backlog - INTEGER
 306        Maximal number of remembered connection requests, which have not
 307        received an acknowledgment from connecting client.
 308        The minimal value is 128 for low memory machines, and it will
 309        increase in proportion to the memory of machine.
 310        If server suffers from overload, try increasing this number.
 311
 312tcp_max_tw_buckets - INTEGER
 313        Maximal number of timewait sockets held by system simultaneously.
 314        If this number is exceeded time-wait socket is immediately destroyed
 315        and warning is printed. This limit exists only to prevent
 316        simple DoS attacks, you _must_ not lower the limit artificially,
 317        but rather increase it (probably, after increasing installed memory),
 318        if network conditions require more than default value.
 319
 320tcp_mem - vector of 3 INTEGERs: min, pressure, max
 321        min: below this number of pages TCP is not bothered about its
 322        memory appetite.
 323
 324        pressure: when amount of memory allocated by TCP exceeds this number
 325        of pages, TCP moderates its memory consumption and enters memory
 326        pressure mode, which is exited when memory consumption falls
 327        under "min".
 328
 329        max: number of pages allowed for queueing by all TCP sockets.
 330
 331        Defaults are calculated at boot time from amount of available
 332        memory.
 333
 334tcp_moderate_rcvbuf - BOOLEAN
 335        If set, TCP performs receive buffer auto-tuning, attempting to
 336        automatically size the buffer (no greater than tcp_rmem[2]) to
 337        match the size required by the path for full throughput.  Enabled by
 338        default.
 339
 340tcp_mtu_probing - INTEGER
 341        Controls TCP Packetization-Layer Path MTU Discovery.  Takes three
 342        values:
 343          0 - Disabled
 344          1 - Disabled by default, enabled when an ICMP black hole detected
 345          2 - Always enabled, use initial MSS of tcp_base_mss.
 346
 347tcp_no_metrics_save - BOOLEAN
 348        By default, TCP saves various connection metrics in the route cache
 349        when the connection closes, so that connections established in the
 350        near future can use these to set initial conditions.  Usually, this
 351        increases overall performance, but may sometimes cause performance
 352        degradation.  If set, TCP will not cache metrics on closing
 353        connections.
 354
 355tcp_orphan_retries - INTEGER
 356        This value influences the timeout of a locally closed TCP connection,
 357        when RTO retransmissions remain unacknowledged.
 358        See tcp_retries2 for more details.
 359
 360        The default value is 8.
 361        If your machine is a loaded WEB server,
 362        you should think about lowering this value, such sockets
 363        may consume significant resources. Cf. tcp_max_orphans.
 364
 365tcp_reordering - INTEGER
 366        Maximal reordering of packets in a TCP stream.
 367        Default: 3
 368
 369tcp_retrans_collapse - BOOLEAN
 370        Bug-to-bug compatibility with some broken printers.
 371        On retransmit try to send bigger packets to work around bugs in
 372        certain TCP stacks.
 373
 374tcp_retries1 - INTEGER
 375        This value influences the time, after which TCP decides, that
 376        something is wrong due to unacknowledged RTO retransmissions,
 377        and reports this suspicion to the network layer.
 378        See tcp_retries2 for more details.
 379
 380        RFC 1122 recommends at least 3 retransmissions, which is the
 381        default.
 382
 383tcp_retries2 - INTEGER
 384        This value influences the timeout of an alive TCP connection,
 385        when RTO retransmissions remain unacknowledged.
 386        Given a value of N, a hypothetical TCP connection following
 387        exponential backoff with an initial RTO of TCP_RTO_MIN would
 388        retransmit N times before killing the connection at the (N+1)th RTO.
 389
 390        The default value of 15 yields a hypothetical timeout of 924.6
 391        seconds and is a lower bound for the effective timeout.
 392        TCP will effectively time out at the first RTO which exceeds the
 393        hypothetical timeout.
 394
 395        RFC 1122 recommends at least 100 seconds for the timeout,
 396        which corresponds to a value of at least 8.
 397
 398tcp_rfc1337 - BOOLEAN
 399        If set, the TCP stack behaves conforming to RFC1337. If unset,
 400        we are not conforming to RFC, but prevent TCP TIME_WAIT
 401        assassination.
 402        Default: 0
 403
 404tcp_rmem - vector of 3 INTEGERs: min, default, max
 405        min: Minimal size of receive buffer used by TCP sockets.
 406        It is guaranteed to each TCP socket, even under moderate memory
 407        pressure.
 408        Default: 1 page
 409
 410        default: initial size of receive buffer used by TCP sockets.
 411        This value overrides net.core.rmem_default used by other protocols.
 412        Default: 87380 bytes. This value results in window of 65535 with
 413        default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
 414        less for default tcp_app_win. See below about these variables.
 415
 416        max: maximal size of receive buffer allowed for automatically
 417        selected receiver buffers for TCP socket. This value does not override
 418        net.core.rmem_max.  Calling setsockopt() with SO_RCVBUF disables
 419        automatic tuning of that socket's receive buffer size, in which
 420        case this value is ignored.
 421        Default: between 87380B and 6MB, depending on RAM size.
 422
 423tcp_sack - BOOLEAN
 424        Enable select acknowledgments (SACKS).
 425
 426tcp_slow_start_after_idle - BOOLEAN
 427        If set, provide RFC2861 behavior and time out the congestion
 428        window after an idle period.  An idle period is defined at
 429        the current RTO.  If unset, the congestion window will not
 430        be timed out after an idle period.
 431        Default: 1
 432
 433tcp_stdurg - BOOLEAN
 434        Use the Host requirements interpretation of the TCP urgent pointer field.
 435        Most hosts use the older BSD interpretation, so if you turn this on
 436        Linux might not communicate correctly with them.
 437        Default: FALSE
 438
 439tcp_synack_retries - INTEGER
 440        Number of times SYNACKs for a passive TCP connection attempt will
 441        be retransmitted. Should not be higher than 255. Default value
 442        is 5, which corresponds to 31seconds till the last retransmission
 443        with the current initial RTO of 1second. With this the final timeout
 444        for a passive TCP connection will happen after 63seconds.
 445
 446tcp_syncookies - BOOLEAN
 447        Only valid when the kernel was compiled with CONFIG_SYNCOOKIES
 448        Send out syncookies when the syn backlog queue of a socket
 449        overflows. This is to prevent against the common 'SYN flood attack'
 450        Default: FALSE
 451
 452        Note, that syncookies is fallback facility.
 453        It MUST NOT be used to help highly loaded servers to stand
 454        against legal connection rate. If you see SYN flood warnings
 455        in your logs, but investigation shows that they occur
 456        because of overload with legal connections, you should tune
 457        another parameters until this warning disappear.
 458        See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
 459
 460        syncookies seriously violate TCP protocol, do not allow
 461        to use TCP extensions, can result in serious degradation
 462        of some services (f.e. SMTP relaying), visible not by you,
 463        but your clients and relays, contacting you. While you see
 464        SYN flood warnings in logs not being really flooded, your server
 465        is seriously misconfigured.
 466
 467tcp_fastopen - INTEGER
 468        Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
 469        in the opening SYN packet. To use this feature, the client application
 470        must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
 471        connect() to perform a TCP handshake automatically.
 472
 473        The values (bitmap) are
 474        1: Enables sending data in the opening SYN on the client.
 475        2: Enables TCP Fast Open on the server side, i.e., allowing data in
 476           a SYN packet to be accepted and passed to the application before
 477           3-way hand shake finishes.
 478        4: Send data in the opening SYN regardless of cookie availability and
 479           without a cookie option.
 480        0x100: Accept SYN data w/o validating the cookie.
 481        0x200: Accept data-in-SYN w/o any cookie option present.
 482        0x400/0x800: Enable Fast Open on all listeners regardless of the
 483           TCP_FASTOPEN socket option. The two different flags designate two
 484           different ways of setting max_qlen without the TCP_FASTOPEN socket
 485           option.
 486
 487        Default: 0
 488
 489        Note that the client & server side Fast Open flags (1 and 2
 490        respectively) must be also enabled before the rest of flags can take
 491        effect.
 492
 493        See include/net/tcp.h and the code for more details.
 494
 495tcp_syn_retries - INTEGER
 496        Number of times initial SYNs for an active TCP connection attempt
 497        will be retransmitted. Should not be higher than 255. Default value
 498        is 6, which corresponds to 63seconds till the last restransmission
 499        with the current initial RTO of 1second. With this the final timeout
 500        for an active TCP connection attempt will happen after 127seconds.
 501
 502tcp_timestamps - BOOLEAN
 503        Enable timestamps as defined in RFC1323.
 504
 505tcp_tso_win_divisor - INTEGER
 506        This allows control over what percentage of the congestion window
 507        can be consumed by a single TSO frame.
 508        The setting of this parameter is a choice between burstiness and
 509        building larger TSO frames.
 510        Default: 3
 511
 512tcp_tw_recycle - BOOLEAN
 513        Enable fast recycling TIME-WAIT sockets. Default value is 0.
 514        It should not be changed without advice/request of technical
 515        experts.
 516
 517tcp_tw_reuse - BOOLEAN
 518        Allow to reuse TIME-WAIT sockets for new connections when it is
 519        safe from protocol viewpoint. Default value is 0.
 520        It should not be changed without advice/request of technical
 521        experts.
 522
 523tcp_window_scaling - BOOLEAN
 524        Enable window scaling as defined in RFC1323.
 525
 526tcp_wmem - vector of 3 INTEGERs: min, default, max
 527        min: Amount of memory reserved for send buffers for TCP sockets.
 528        Each TCP socket has rights to use it due to fact of its birth.
 529        Default: 1 page
 530
 531        default: initial size of send buffer used by TCP sockets.  This
 532        value overrides net.core.wmem_default used by other protocols.
 533        It is usually lower than net.core.wmem_default.
 534        Default: 16K
 535
 536        max: Maximal amount of memory allowed for automatically tuned
 537        send buffers for TCP sockets. This value does not override
 538        net.core.wmem_max.  Calling setsockopt() with SO_SNDBUF disables
 539        automatic tuning of that socket's send buffer size, in which case
 540        this value is ignored.
 541        Default: between 64K and 4MB, depending on RAM size.
 542
 543tcp_workaround_signed_windows - BOOLEAN
 544        If set, assume no receipt of a window scaling option means the
 545        remote TCP is broken and treats the window as a signed quantity.
 546        If unset, assume the remote TCP is not broken even if we do
 547        not receive a window scaling option from them.
 548        Default: 0
 549
 550tcp_dma_copybreak - INTEGER
 551        Lower limit, in bytes, of the size of socket reads that will be
 552        offloaded to a DMA copy engine, if one is present in the system
 553        and CONFIG_NET_DMA is enabled.
 554        Default: 4096
 555
 556tcp_thin_linear_timeouts - BOOLEAN
 557        Enable dynamic triggering of linear timeouts for thin streams.
 558        If set, a check is performed upon retransmission by timeout to
 559        determine if the stream is thin (less than 4 packets in flight).
 560        As long as the stream is found to be thin, up to 6 linear
 561        timeouts may be performed before exponential backoff mode is
 562        initiated. This improves retransmission latency for
 563        non-aggressive thin streams, often found to be time-dependent.
 564        For more information on thin streams, see
 565        Documentation/networking/tcp-thin.txt
 566        Default: 0
 567
 568tcp_thin_dupack - BOOLEAN
 569        Enable dynamic triggering of retransmissions after one dupACK
 570        for thin streams. If set, a check is performed upon reception
 571        of a dupACK to determine if the stream is thin (less than 4
 572        packets in flight). As long as the stream is found to be thin,
 573        data is retransmitted on the first received dupACK. This
 574        improves retransmission latency for non-aggressive thin
 575        streams, often found to be time-dependent.
 576        For more information on thin streams, see
 577        Documentation/networking/tcp-thin.txt
 578        Default: 0
 579
 580tcp_limit_output_bytes - INTEGER
 581        Controls TCP Small Queue limit per tcp socket.
 582        TCP bulk sender tends to increase packets in flight until it
 583        gets losses notifications. With SNDBUF autotuning, this can
 584        result in a large amount of packets queued in qdisc/device
 585        on the local machine, hurting latency of other flows, for
 586        typical pfifo_fast qdiscs.
 587        tcp_limit_output_bytes limits the number of bytes on qdisc
 588        or device to reduce artificial RTT/cwnd and reduce bufferbloat.
 589        Note: For GSO/TSO enabled flows, we try to have at least two
 590        packets in flight. Reducing tcp_limit_output_bytes might also
 591        reduce the size of individual GSO packet (64KB being the max)
 592        Default: 131072
 593
 594tcp_challenge_ack_limit - INTEGER
 595        Limits number of Challenge ACK sent per second, as recommended
 596        in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
 597        Default: 100
 598
 599UDP variables:
 600
 601udp_mem - vector of 3 INTEGERs: min, pressure, max
 602        Number of pages allowed for queueing by all UDP sockets.
 603
 604        min: Below this number of pages UDP is not bothered about its
 605        memory appetite. When amount of memory allocated by UDP exceeds
 606        this number, UDP starts to moderate memory usage.
 607
 608        pressure: This value was introduced to follow format of tcp_mem.
 609
 610        max: Number of pages allowed for queueing by all UDP sockets.
 611
 612        Default is calculated at boot time from amount of available memory.
 613
 614udp_rmem_min - INTEGER
 615        Minimal size of receive buffer used by UDP sockets in moderation.
 616        Each UDP socket is able to use the size for receiving data, even if
 617        total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
 618        Default: 1 page
 619
 620udp_wmem_min - INTEGER
 621        Minimal size of send buffer used by UDP sockets in moderation.
 622        Each UDP socket is able to use the size for sending data, even if
 623        total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
 624        Default: 1 page
 625
 626CIPSOv4 Variables:
 627
 628cipso_cache_enable - BOOLEAN
 629        If set, enable additions to and lookups from the CIPSO label mapping
 630        cache.  If unset, additions are ignored and lookups always result in a
 631        miss.  However, regardless of the setting the cache is still
 632        invalidated when required when means you can safely toggle this on and
 633        off and the cache will always be "safe".
 634        Default: 1
 635
 636cipso_cache_bucket_size - INTEGER
 637        The CIPSO label cache consists of a fixed size hash table with each
 638        hash bucket containing a number of cache entries.  This variable limits
 639        the number of entries in each hash bucket; the larger the value the
 640        more CIPSO label mappings that can be cached.  When the number of
 641        entries in a given hash bucket reaches this limit adding new entries
 642        causes the oldest entry in the bucket to be removed to make room.
 643        Default: 10
 644
 645cipso_rbm_optfmt - BOOLEAN
 646        Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
 647        the CIPSO draft specification (see Documentation/netlabel for details).
 648        This means that when set the CIPSO tag will be padded with empty
 649        categories in order to make the packet data 32-bit aligned.
 650        Default: 0
 651
 652cipso_rbm_structvalid - BOOLEAN
 653        If set, do a very strict check of the CIPSO option when
 654        ip_options_compile() is called.  If unset, relax the checks done during
 655        ip_options_compile().  Either way is "safe" as errors are caught else
 656        where in the CIPSO processing code but setting this to 0 (False) should
 657        result in less work (i.e. it should be faster) but could cause problems
 658        with other implementations that require strict checking.
 659        Default: 0
 660
 661IP Variables:
 662
 663ip_local_port_range - 2 INTEGERS
 664        Defines the local port range that is used by TCP and UDP to
 665        choose the local port. The first number is the first, the
 666        second the last local port number. The default values are
 667        32768 and 61000 respectively.
 668
 669ip_local_reserved_ports - list of comma separated ranges
 670        Specify the ports which are reserved for known third-party
 671        applications. These ports will not be used by automatic port
 672        assignments (e.g. when calling connect() or bind() with port
 673        number 0). Explicit port allocation behavior is unchanged.
 674
 675        The format used for both input and output is a comma separated
 676        list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
 677        10). Writing to the file will clear all previously reserved
 678        ports and update the current list with the one given in the
 679        input.
 680
 681        Note that ip_local_port_range and ip_local_reserved_ports
 682        settings are independent and both are considered by the kernel
 683        when determining which ports are available for automatic port
 684        assignments.
 685
 686        You can reserve ports which are not in the current
 687        ip_local_port_range, e.g.:
 688
 689        $ cat /proc/sys/net/ipv4/ip_local_port_range
 690        32000   61000
 691        $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
 692        8080,9148
 693
 694        although this is redundant. However such a setting is useful
 695        if later the port range is changed to a value that will
 696        include the reserved ports.
 697
 698        Default: Empty
 699
 700ip_nonlocal_bind - BOOLEAN
 701        If set, allows processes to bind() to non-local IP addresses,
 702        which can be quite useful - but may break some applications.
 703        Default: 0
 704
 705ip_dynaddr - BOOLEAN
 706        If set non-zero, enables support for dynamic addresses.
 707        If set to a non-zero value larger than 1, a kernel log
 708        message will be printed when dynamic address rewriting
 709        occurs.
 710        Default: 0
 711
 712icmp_echo_ignore_all - BOOLEAN
 713        If set non-zero, then the kernel will ignore all ICMP ECHO
 714        requests sent to it.
 715        Default: 0
 716
 717icmp_echo_ignore_broadcasts - BOOLEAN
 718        If set non-zero, then the kernel will ignore all ICMP ECHO and
 719        TIMESTAMP requests sent to it via broadcast/multicast.
 720        Default: 1
 721
 722icmp_ratelimit - INTEGER
 723        Limit the maximal rates for sending ICMP packets whose type matches
 724        icmp_ratemask (see below) to specific targets.
 725        0 to disable any limiting,
 726        otherwise the minimal space between responses in milliseconds.
 727        Default: 1000
 728
 729icmp_ratemask - INTEGER
 730        Mask made of ICMP types for which rates are being limited.
 731        Significant bits: IHGFEDCBA9876543210
 732        Default mask:     0000001100000011000 (6168)
 733
 734        Bit definitions (see include/linux/icmp.h):
 735                0 Echo Reply
 736                3 Destination Unreachable *
 737                4 Source Quench *
 738                5 Redirect
 739                8 Echo Request
 740                B Time Exceeded *
 741                C Parameter Problem *
 742                D Timestamp Request
 743                E Timestamp Reply
 744                F Info Request
 745                G Info Reply
 746                H Address Mask Request
 747                I Address Mask Reply
 748
 749        * These are rate limited by default (see default mask above)
 750
 751icmp_ignore_bogus_error_responses - BOOLEAN
 752        Some routers violate RFC1122 by sending bogus responses to broadcast
 753        frames.  Such violations are normally logged via a kernel warning.
 754        If this is set to TRUE, the kernel will not give such warnings, which
 755        will avoid log file clutter.
 756        Default: FALSE
 757
 758icmp_errors_use_inbound_ifaddr - BOOLEAN
 759
 760        If zero, icmp error messages are sent with the primary address of
 761        the exiting interface.
 762
 763        If non-zero, the message will be sent with the primary address of
 764        the interface that received the packet that caused the icmp error.
 765        This is the behaviour network many administrators will expect from
 766        a router. And it can make debugging complicated network layouts
 767        much easier.
 768
 769        Note that if no primary address exists for the interface selected,
 770        then the primary address of the first non-loopback interface that
 771        has one will be used regardless of this setting.
 772
 773        Default: 0
 774
 775igmp_max_memberships - INTEGER
 776        Change the maximum number of multicast groups we can subscribe to.
 777        Default: 20
 778
 779        Theoretical maximum value is bounded by having to send a membership
 780        report in a single datagram (i.e. the report can't span multiple
 781        datagrams, or risk confusing the switch and leaving groups you don't
 782        intend to).
 783
 784        The number of supported groups 'M' is bounded by the number of group
 785        report entries you can fit into a single datagram of 65535 bytes.
 786
 787        M = 65536-sizeof (ip header)/(sizeof(Group record))
 788
 789        Group records are variable length, with a minimum of 12 bytes.
 790        So net.ipv4.igmp_max_memberships should not be set higher than:
 791
 792        (65536-24) / 12 = 5459
 793
 794        The value 5459 assumes no IP header options, so in practice
 795        this number may be lower.
 796
 797        conf/interface/*  changes special settings per interface (where
 798        "interface" is the name of your network interface)
 799
 800        conf/all/*        is special, changes the settings for all interfaces
 801
 802log_martians - BOOLEAN
 803        Log packets with impossible addresses to kernel log.
 804        log_martians for the interface will be enabled if at least one of
 805        conf/{all,interface}/log_martians is set to TRUE,
 806        it will be disabled otherwise
 807
 808accept_redirects - BOOLEAN
 809        Accept ICMP redirect messages.
 810        accept_redirects for the interface will be enabled if:
 811        - both conf/{all,interface}/accept_redirects are TRUE in the case
 812          forwarding for the interface is enabled
 813        or
 814        - at least one of conf/{all,interface}/accept_redirects is TRUE in the
 815          case forwarding for the interface is disabled
 816        accept_redirects for the interface will be disabled otherwise
 817        default TRUE (host)
 818                FALSE (router)
 819
 820forwarding - BOOLEAN
 821        Enable IP forwarding on this interface.
 822
 823mc_forwarding - BOOLEAN
 824        Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
 825        and a multicast routing daemon is required.
 826        conf/all/mc_forwarding must also be set to TRUE to enable multicast
 827        routing for the interface
 828
 829medium_id - INTEGER
 830        Integer value used to differentiate the devices by the medium they
 831        are attached to. Two devices can have different id values when
 832        the broadcast packets are received only on one of them.
 833        The default value 0 means that the device is the only interface
 834        to its medium, value of -1 means that medium is not known.
 835
 836        Currently, it is used to change the proxy_arp behavior:
 837        the proxy_arp feature is enabled for packets forwarded between
 838        two devices attached to different media.
 839
 840proxy_arp - BOOLEAN
 841        Do proxy arp.
 842        proxy_arp for the interface will be enabled if at least one of
 843        conf/{all,interface}/proxy_arp is set to TRUE,
 844        it will be disabled otherwise
 845
 846proxy_arp_pvlan - BOOLEAN
 847        Private VLAN proxy arp.
 848        Basically allow proxy arp replies back to the same interface
 849        (from which the ARP request/solicitation was received).
 850
 851        This is done to support (ethernet) switch features, like RFC
 852        3069, where the individual ports are NOT allowed to
 853        communicate with each other, but they are allowed to talk to
 854        the upstream router.  As described in RFC 3069, it is possible
 855        to allow these hosts to communicate through the upstream
 856        router by proxy_arp'ing. Don't need to be used together with
 857        proxy_arp.
 858
 859        This technology is known by different names:
 860          In RFC 3069 it is called VLAN Aggregation.
 861          Cisco and Allied Telesyn call it Private VLAN.
 862          Hewlett-Packard call it Source-Port filtering or port-isolation.
 863          Ericsson call it MAC-Forced Forwarding (RFC Draft).
 864
 865shared_media - BOOLEAN
 866        Send(router) or accept(host) RFC1620 shared media redirects.
 867        Overrides ip_secure_redirects.
 868        shared_media for the interface will be enabled if at least one of
 869        conf/{all,interface}/shared_media is set to TRUE,
 870        it will be disabled otherwise
 871        default TRUE
 872
 873secure_redirects - BOOLEAN
 874        Accept ICMP redirect messages only for gateways,
 875        listed in default gateway list.
 876        secure_redirects for the interface will be enabled if at least one of
 877        conf/{all,interface}/secure_redirects is set to TRUE,
 878        it will be disabled otherwise
 879        default TRUE
 880
 881send_redirects - BOOLEAN
 882        Send redirects, if router.
 883        send_redirects for the interface will be enabled if at least one of
 884        conf/{all,interface}/send_redirects is set to TRUE,
 885        it will be disabled otherwise
 886        Default: TRUE
 887
 888bootp_relay - BOOLEAN
 889        Accept packets with source address 0.b.c.d destined
 890        not to this host as local ones. It is supposed, that
 891        BOOTP relay daemon will catch and forward such packets.
 892        conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
 893        for the interface
 894        default FALSE
 895        Not Implemented Yet.
 896
 897accept_source_route - BOOLEAN
 898        Accept packets with SRR option.
 899        conf/all/accept_source_route must also be set to TRUE to accept packets
 900        with SRR option on the interface
 901        default TRUE (router)
 902                FALSE (host)
 903
 904accept_local - BOOLEAN
 905        Accept packets with local source addresses. In combination
 906        with suitable routing, this can be used to direct packets
 907        between two local interfaces over the wire and have them
 908        accepted properly.
 909
 910        rp_filter must be set to a non-zero value in order for
 911        accept_local to have an effect.
 912
 913        default FALSE
 914
 915route_localnet - BOOLEAN
 916        Do not consider loopback addresses as martian source or destination
 917        while routing. This enables the use of 127/8 for local routing purposes.
 918        default FALSE
 919
 920rp_filter - INTEGER
 921        0 - No source validation.
 922        1 - Strict mode as defined in RFC3704 Strict Reverse Path
 923            Each incoming packet is tested against the FIB and if the interface
 924            is not the best reverse path the packet check will fail.
 925            By default failed packets are discarded.
 926        2 - Loose mode as defined in RFC3704 Loose Reverse Path
 927            Each incoming packet's source address is also tested against the FIB
 928            and if the source address is not reachable via any interface
 929            the packet check will fail.
 930
 931        Current recommended practice in RFC3704 is to enable strict mode
 932        to prevent IP spoofing from DDos attacks. If using asymmetric routing
 933        or other complicated routing, then loose mode is recommended.
 934
 935        The max value from conf/{all,interface}/rp_filter is used
 936        when doing source validation on the {interface}.
 937
 938        Default value is 0. Note that some distributions enable it
 939        in startup scripts.
 940
 941arp_filter - BOOLEAN
 942        1 - Allows you to have multiple network interfaces on the same
 943        subnet, and have the ARPs for each interface be answered
 944        based on whether or not the kernel would route a packet from
 945        the ARP'd IP out that interface (therefore you must use source
 946        based routing for this to work). In other words it allows control
 947        of which cards (usually 1) will respond to an arp request.
 948
 949        0 - (default) The kernel can respond to arp requests with addresses
 950        from other interfaces. This may seem wrong but it usually makes
 951        sense, because it increases the chance of successful communication.
 952        IP addresses are owned by the complete host on Linux, not by
 953        particular interfaces. Only for more complex setups like load-
 954        balancing, does this behaviour cause problems.
 955
 956        arp_filter for the interface will be enabled if at least one of
 957        conf/{all,interface}/arp_filter is set to TRUE,
 958        it will be disabled otherwise
 959
 960arp_announce - INTEGER
 961        Define different restriction levels for announcing the local
 962        source IP address from IP packets in ARP requests sent on
 963        interface:
 964        0 - (default) Use any local address, configured on any interface
 965        1 - Try to avoid local addresses that are not in the target's
 966        subnet for this interface. This mode is useful when target
 967        hosts reachable via this interface require the source IP
 968        address in ARP requests to be part of their logical network
 969        configured on the receiving interface. When we generate the
 970        request we will check all our subnets that include the
 971        target IP and will preserve the source address if it is from
 972        such subnet. If there is no such subnet we select source
 973        address according to the rules for level 2.
 974        2 - Always use the best local address for this target.
 975        In this mode we ignore the source address in the IP packet
 976        and try to select local address that we prefer for talks with
 977        the target host. Such local address is selected by looking
 978        for primary IP addresses on all our subnets on the outgoing
 979        interface that include the target IP address. If no suitable
 980        local address is found we select the first local address
 981        we have on the outgoing interface or on all other interfaces,
 982        with the hope we will receive reply for our request and
 983        even sometimes no matter the source IP address we announce.
 984
 985        The max value from conf/{all,interface}/arp_announce is used.
 986
 987        Increasing the restriction level gives more chance for
 988        receiving answer from the resolved target while decreasing
 989        the level announces more valid sender's information.
 990
 991arp_ignore - INTEGER
 992        Define different modes for sending replies in response to
 993        received ARP requests that resolve local target IP addresses:
 994        0 - (default): reply for any local target IP address, configured
 995        on any interface
 996        1 - reply only if the target IP address is local address
 997        configured on the incoming interface
 998        2 - reply only if the target IP address is local address
 999        configured on the incoming interface and both with the
1000        sender's IP address are part from same subnet on this interface
1001        3 - do not reply for local addresses configured with scope host,
1002        only resolutions for global and link addresses are replied
1003        4-7 - reserved
1004        8 - do not reply for all local addresses
1005
1006        The max value from conf/{all,interface}/arp_ignore is used
1007        when ARP request is received on the {interface}
1008
1009arp_notify - BOOLEAN
1010        Define mode for notification of address and device changes.
1011        0 - (default): do nothing
1012        1 - Generate gratuitous arp requests when device is brought up
1013            or hardware address changes.
1014
1015arp_accept - BOOLEAN
1016        Define behavior for gratuitous ARP frames who's IP is not
1017        already present in the ARP table:
1018        0 - don't create new entries in the ARP table
1019        1 - create new entries in the ARP table
1020
1021        Both replies and requests type gratuitous arp will trigger the
1022        ARP table to be updated, if this setting is on.
1023
1024        If the ARP table already contains the IP address of the
1025        gratuitous arp frame, the arp table will be updated regardless
1026        if this setting is on or off.
1027
1028
1029app_solicit - INTEGER
1030        The maximum number of probes to send to the user space ARP daemon
1031        via netlink before dropping back to multicast probes (see
1032        mcast_solicit).  Defaults to 0.
1033
1034disable_policy - BOOLEAN
1035        Disable IPSEC policy (SPD) for this interface
1036
1037disable_xfrm - BOOLEAN
1038        Disable IPSEC encryption on this interface, whatever the policy
1039
1040
1041
1042tag - INTEGER
1043        Allows you to write a number, which can be used as required.
1044        Default value is 0.
1045
1046Alexey Kuznetsov.
1047kuznet@ms2.inr.ac.ru
1048
1049Updated by:
1050Andi Kleen
1051ak@muc.de
1052Nicolas Delon
1053delon.nicolas@wanadoo.fr
1054
1055
1056
1057
1058/proc/sys/net/ipv6/* Variables:
1059
1060IPv6 has no global variables such as tcp_*.  tcp_* settings under ipv4/ also
1061apply to IPv6 [XXX?].
1062
1063bindv6only - BOOLEAN
1064        Default value for IPV6_V6ONLY socket option,
1065        which restricts use of the IPv6 socket to IPv6 communication
1066        only.
1067                TRUE: disable IPv4-mapped address feature
1068                FALSE: enable IPv4-mapped address feature
1069
1070        Default: FALSE (as specified in RFC3493)
1071
1072IPv6 Fragmentation:
1073
1074ip6frag_high_thresh - INTEGER
1075        Maximum memory used to reassemble IPv6 fragments. When
1076        ip6frag_high_thresh bytes of memory is allocated for this purpose,
1077        the fragment handler will toss packets until ip6frag_low_thresh
1078        is reached.
1079
1080ip6frag_low_thresh - INTEGER
1081        See ip6frag_high_thresh
1082
1083ip6frag_time - INTEGER
1084        Time in seconds to keep an IPv6 fragment in memory.
1085
1086ip6frag_secret_interval - INTEGER
1087        Regeneration interval (in seconds) of the hash secret (or lifetime
1088        for the hash secret) for IPv6 fragments.
1089        Default: 600
1090
1091conf/default/*:
1092        Change the interface-specific default settings.
1093
1094
1095conf/all/*:
1096        Change all the interface-specific settings.
1097
1098        [XXX:  Other special features than forwarding?]
1099
1100conf/all/forwarding - BOOLEAN
1101        Enable global IPv6 forwarding between all interfaces.
1102
1103        IPv4 and IPv6 work differently here; e.g. netfilter must be used
1104        to control which interfaces may forward packets and which not.
1105
1106        This also sets all interfaces' Host/Router setting
1107        'forwarding' to the specified value.  See below for details.
1108
1109        This referred to as global forwarding.
1110
1111proxy_ndp - BOOLEAN
1112        Do proxy ndp.
1113
1114conf/interface/*:
1115        Change special settings per interface.
1116
1117        The functional behaviour for certain settings is different
1118        depending on whether local forwarding is enabled or not.
1119
1120accept_ra - INTEGER
1121        Accept Router Advertisements; autoconfigure using them.
1122
1123        It also determines whether or not to transmit Router
1124        Solicitations. If and only if the functional setting is to
1125        accept Router Advertisements, Router Solicitations will be
1126        transmitted.
1127
1128        Possible values are:
1129                0 Do not accept Router Advertisements.
1130                1 Accept Router Advertisements if forwarding is disabled.
1131                2 Overrule forwarding behaviour. Accept Router Advertisements
1132                  even if forwarding is enabled.
1133
1134        Functional default: enabled if local forwarding is disabled.
1135                            disabled if local forwarding is enabled.
1136
1137accept_ra_defrtr - BOOLEAN
1138        Learn default router in Router Advertisement.
1139
1140        Functional default: enabled if accept_ra is enabled.
1141                            disabled if accept_ra is disabled.
1142
1143accept_ra_pinfo - BOOLEAN
1144        Learn Prefix Information in Router Advertisement.
1145
1146        Functional default: enabled if accept_ra is enabled.
1147                            disabled if accept_ra is disabled.
1148
1149accept_ra_rt_info_max_plen - INTEGER
1150        Maximum prefix length of Route Information in RA.
1151
1152        Route Information w/ prefix larger than or equal to this
1153        variable shall be ignored.
1154
1155        Functional default: 0 if accept_ra_rtr_pref is enabled.
1156                            -1 if accept_ra_rtr_pref is disabled.
1157
1158accept_ra_rtr_pref - BOOLEAN
1159        Accept Router Preference in RA.
1160
1161        Functional default: enabled if accept_ra is enabled.
1162                            disabled if accept_ra is disabled.
1163
1164accept_redirects - BOOLEAN
1165        Accept Redirects.
1166
1167        Functional default: enabled if local forwarding is disabled.
1168                            disabled if local forwarding is enabled.
1169
1170accept_source_route - INTEGER
1171        Accept source routing (routing extension header).
1172
1173        >= 0: Accept only routing header type 2.
1174        < 0: Do not accept routing header.
1175
1176        Default: 0
1177
1178autoconf - BOOLEAN
1179        Autoconfigure addresses using Prefix Information in Router
1180        Advertisements.
1181
1182        Functional default: enabled if accept_ra_pinfo is enabled.
1183                            disabled if accept_ra_pinfo is disabled.
1184
1185dad_transmits - INTEGER
1186        The amount of Duplicate Address Detection probes to send.
1187        Default: 1
1188
1189forwarding - INTEGER
1190        Configure interface-specific Host/Router behaviour.
1191
1192        Note: It is recommended to have the same setting on all
1193        interfaces; mixed router/host scenarios are rather uncommon.
1194
1195        Possible values are:
1196                0 Forwarding disabled
1197                1 Forwarding enabled
1198
1199        FALSE (0):
1200
1201        By default, Host behaviour is assumed.  This means:
1202
1203        1. IsRouter flag is not set in Neighbour Advertisements.
1204        2. If accept_ra is TRUE (default), transmit Router
1205           Solicitations.
1206        3. If accept_ra is TRUE (default), accept Router
1207           Advertisements (and do autoconfiguration).
1208        4. If accept_redirects is TRUE (default), accept Redirects.
1209
1210        TRUE (1):
1211
1212        If local forwarding is enabled, Router behaviour is assumed.
1213        This means exactly the reverse from the above:
1214
1215        1. IsRouter flag is set in Neighbour Advertisements.
1216        2. Router Solicitations are not sent unless accept_ra is 2.
1217        3. Router Advertisements are ignored unless accept_ra is 2.
1218        4. Redirects are ignored.
1219
1220        Default: 0 (disabled) if global forwarding is disabled (default),
1221                 otherwise 1 (enabled).
1222
1223hop_limit - INTEGER
1224        Default Hop Limit to set.
1225        Default: 64
1226
1227mtu - INTEGER
1228        Default Maximum Transfer Unit
1229        Default: 1280 (IPv6 required minimum)
1230
1231router_probe_interval - INTEGER
1232        Minimum interval (in seconds) between Router Probing described
1233        in RFC4191.
1234
1235        Default: 60
1236
1237router_solicitation_delay - INTEGER
1238        Number of seconds to wait after interface is brought up
1239        before sending Router Solicitations.
1240        Default: 1
1241
1242router_solicitation_interval - INTEGER
1243        Number of seconds to wait between Router Solicitations.
1244        Default: 4
1245
1246router_solicitations - INTEGER
1247        Number of Router Solicitations to send until assuming no
1248        routers are present.
1249        Default: 3
1250
1251use_tempaddr - INTEGER
1252        Preference for Privacy Extensions (RFC3041).
1253          <= 0 : disable Privacy Extensions
1254          == 1 : enable Privacy Extensions, but prefer public
1255                 addresses over temporary addresses.
1256          >  1 : enable Privacy Extensions and prefer temporary
1257                 addresses over public addresses.
1258        Default:  0 (for most devices)
1259                 -1 (for point-to-point devices and loopback devices)
1260
1261temp_valid_lft - INTEGER
1262        valid lifetime (in seconds) for temporary addresses.
1263        Default: 604800 (7 days)
1264
1265temp_prefered_lft - INTEGER
1266        Preferred lifetime (in seconds) for temporary addresses.
1267        Default: 86400 (1 day)
1268
1269max_desync_factor - INTEGER
1270        Maximum value for DESYNC_FACTOR, which is a random value
1271        that ensures that clients don't synchronize with each
1272        other and generate new addresses at exactly the same time.
1273        value is in seconds.
1274        Default: 600
1275
1276regen_max_retry - INTEGER
1277        Number of attempts before give up attempting to generate
1278        valid temporary addresses.
1279        Default: 5
1280
1281max_addresses - INTEGER
1282        Maximum number of autoconfigured addresses per interface.  Setting
1283        to zero disables the limitation.  It is not recommended to set this
1284        value too large (or to zero) because it would be an easy way to
1285        crash the kernel by allowing too many addresses to be created.
1286        Default: 16
1287
1288disable_ipv6 - BOOLEAN
1289        Disable IPv6 operation.  If accept_dad is set to 2, this value
1290        will be dynamically set to TRUE if DAD fails for the link-local
1291        address.
1292        Default: FALSE (enable IPv6 operation)
1293
1294        When this value is changed from 1 to 0 (IPv6 is being enabled),
1295        it will dynamically create a link-local address on the given
1296        interface and start Duplicate Address Detection, if necessary.
1297
1298        When this value is changed from 0 to 1 (IPv6 is being disabled),
1299        it will dynamically delete all address on the given interface.
1300
1301accept_dad - INTEGER
1302        Whether to accept DAD (Duplicate Address Detection).
1303        0: Disable DAD
1304        1: Enable DAD (default)
1305        2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1306           link-local address has been found.
1307
1308force_tllao - BOOLEAN
1309        Enable sending the target link-layer address option even when
1310        responding to a unicast neighbor solicitation.
1311        Default: FALSE
1312
1313        Quoting from RFC 2461, section 4.4, Target link-layer address:
1314
1315        "The option MUST be included for multicast solicitations in order to
1316        avoid infinite Neighbor Solicitation "recursion" when the peer node
1317        does not have a cache entry to return a Neighbor Advertisements
1318        message.  When responding to unicast solicitations, the option can be
1319        omitted since the sender of the solicitation has the correct link-
1320        layer address; otherwise it would not have be able to send the unicast
1321        solicitation in the first place. However, including the link-layer
1322        address in this case adds little overhead and eliminates a potential
1323        race condition where the sender deletes the cached link-layer address
1324        prior to receiving a response to a previous solicitation."
1325
1326icmp/*:
1327ratelimit - INTEGER
1328        Limit the maximal rates for sending ICMPv6 packets.
1329        0 to disable any limiting,
1330        otherwise the minimal space between responses in milliseconds.
1331        Default: 1000
1332
1333
1334IPv6 Update by:
1335Pekka Savola <pekkas@netcore.fi>
1336YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1337
1338
1339/proc/sys/net/bridge/* Variables:
1340
1341bridge-nf-call-arptables - BOOLEAN
1342        1 : pass bridged ARP traffic to arptables' FORWARD chain.
1343        0 : disable this.
1344        Default: 1
1345
1346bridge-nf-call-iptables - BOOLEAN
1347        1 : pass bridged IPv4 traffic to iptables' chains.
1348        0 : disable this.
1349        Default: 1
1350
1351bridge-nf-call-ip6tables - BOOLEAN
1352        1 : pass bridged IPv6 traffic to ip6tables' chains.
1353        0 : disable this.
1354        Default: 1
1355
1356bridge-nf-filter-vlan-tagged - BOOLEAN
1357        1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1358        0 : disable this.
1359        Default: 0
1360
1361bridge-nf-filter-pppoe-tagged - BOOLEAN
1362        1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1363        0 : disable this.
1364        Default: 0
1365
1366bridge-nf-pass-vlan-input-dev - BOOLEAN
1367        1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1368        interface on the bridge and set the netfilter input device to the vlan.
1369        This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1370        target work with vlan-on-top-of-bridge interfaces.  When no matching
1371        vlan interface is found, or this switch is off, the input device is
1372        set to the bridge interface.
1373        0: disable bridge netfilter vlan interface lookup.
1374        Default: 0
1375
1376proc/sys/net/sctp/* Variables:
1377
1378addip_enable - BOOLEAN
1379        Enable or disable extension of  Dynamic Address Reconfiguration
1380        (ADD-IP) functionality specified in RFC5061.  This extension provides
1381        the ability to dynamically add and remove new addresses for the SCTP
1382        associations.
1383
1384        1: Enable extension.
1385
1386        0: Disable extension.
1387
1388        Default: 0
1389
1390addip_noauth_enable - BOOLEAN
1391        Dynamic Address Reconfiguration (ADD-IP) requires the use of
1392        authentication to protect the operations of adding or removing new
1393        addresses.  This requirement is mandated so that unauthorized hosts
1394        would not be able to hijack associations.  However, older
1395        implementations may not have implemented this requirement while
1396        allowing the ADD-IP extension.  For reasons of interoperability,
1397        we provide this variable to control the enforcement of the
1398        authentication requirement.
1399
1400        1: Allow ADD-IP extension to be used without authentication.  This
1401           should only be set in a closed environment for interoperability
1402           with older implementations.
1403
1404        0: Enforce the authentication requirement
1405
1406        Default: 0
1407
1408auth_enable - BOOLEAN
1409        Enable or disable Authenticated Chunks extension.  This extension
1410        provides the ability to send and receive authenticated chunks and is
1411        required for secure operation of Dynamic Address Reconfiguration
1412        (ADD-IP) extension.
1413
1414        1: Enable this extension.
1415        0: Disable this extension.
1416
1417        Default: 0
1418
1419prsctp_enable - BOOLEAN
1420        Enable or disable the Partial Reliability extension (RFC3758) which
1421        is used to notify peers that a given DATA should no longer be expected.
1422
1423        1: Enable extension
1424        0: Disable
1425
1426        Default: 1
1427
1428max_burst - INTEGER
1429        The limit of the number of new packets that can be initially sent.  It
1430        controls how bursty the generated traffic can be.
1431
1432        Default: 4
1433
1434association_max_retrans - INTEGER
1435        Set the maximum number for retransmissions that an association can
1436        attempt deciding that the remote end is unreachable.  If this value
1437        is exceeded, the association is terminated.
1438
1439        Default: 10
1440
1441max_init_retransmits - INTEGER
1442        The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1443        that an association will attempt before declaring the destination
1444        unreachable and terminating.
1445
1446        Default: 8
1447
1448path_max_retrans - INTEGER
1449        The maximum number of retransmissions that will be attempted on a given
1450        path.  Once this threshold is exceeded, the path is considered
1451        unreachable, and new traffic will use a different path when the
1452        association is multihomed.
1453
1454        Default: 5
1455
1456pf_retrans - INTEGER
1457        The number of retransmissions that will be attempted on a given path
1458        before traffic is redirected to an alternate transport (should one
1459        exist).  Note this is distinct from path_max_retrans, as a path that
1460        passes the pf_retrans threshold can still be used.  Its only
1461        deprioritized when a transmission path is selected by the stack.  This
1462        setting is primarily used to enable fast failover mechanisms without
1463        having to reduce path_max_retrans to a very low value.  See:
1464        http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1465        for details.  Note also that a value of pf_retrans > path_max_retrans
1466        disables this feature
1467
1468        Default: 0
1469
1470rto_initial - INTEGER
1471        The initial round trip timeout value in milliseconds that will be used
1472        in calculating round trip times.  This is the initial time interval
1473        for retransmissions.
1474
1475        Default: 3000
1476
1477rto_max - INTEGER
1478        The maximum value (in milliseconds) of the round trip timeout.  This
1479        is the largest time interval that can elapse between retransmissions.
1480
1481        Default: 60000
1482
1483rto_min - INTEGER
1484        The minimum value (in milliseconds) of the round trip timeout.  This
1485        is the smallest time interval the can elapse between retransmissions.
1486
1487        Default: 1000
1488
1489hb_interval - INTEGER
1490        The interval (in milliseconds) between HEARTBEAT chunks.  These chunks
1491        are sent at the specified interval on idle paths to probe the state of
1492        a given path between 2 associations.
1493
1494        Default: 30000
1495
1496sack_timeout - INTEGER
1497        The amount of time (in milliseconds) that the implementation will wait
1498        to send a SACK.
1499
1500        Default: 200
1501
1502valid_cookie_life - INTEGER
1503        The default lifetime of the SCTP cookie (in milliseconds).  The cookie
1504        is used during association establishment.
1505
1506        Default: 60000
1507
1508cookie_preserve_enable - BOOLEAN
1509        Enable or disable the ability to extend the lifetime of the SCTP cookie
1510        that is used during the establishment phase of SCTP association
1511
1512        1: Enable cookie lifetime extension.
1513        0: Disable
1514
1515        Default: 1
1516
1517rcvbuf_policy - INTEGER
1518        Determines if the receive buffer is attributed to the socket or to
1519        association.   SCTP supports the capability to create multiple
1520        associations on a single socket.  When using this capability, it is
1521        possible that a single stalled association that's buffering a lot
1522        of data may block other associations from delivering their data by
1523        consuming all of the receive buffer space.  To work around this,
1524        the rcvbuf_policy could be set to attribute the receiver buffer space
1525        to each association instead of the socket.  This prevents the described
1526        blocking.
1527
1528        1: rcvbuf space is per association
1529        0: recbuf space is per socket
1530
1531        Default: 0
1532
1533sndbuf_policy - INTEGER
1534        Similar to rcvbuf_policy above, this applies to send buffer space.
1535
1536        1: Send buffer is tracked per association
1537        0: Send buffer is tracked per socket.
1538
1539        Default: 0
1540
1541sctp_mem - vector of 3 INTEGERs: min, pressure, max
1542        Number of pages allowed for queueing by all SCTP sockets.
1543
1544        min: Below this number of pages SCTP is not bothered about its
1545        memory appetite. When amount of memory allocated by SCTP exceeds
1546        this number, SCTP starts to moderate memory usage.
1547
1548        pressure: This value was introduced to follow format of tcp_mem.
1549
1550        max: Number of pages allowed for queueing by all SCTP sockets.
1551
1552        Default is calculated at boot time from amount of available memory.
1553
1554sctp_rmem - vector of 3 INTEGERs: min, default, max
1555        Only the first value ("min") is used, "default" and "max" are
1556        ignored.
1557
1558        min: Minimal size of receive buffer used by SCTP socket.
1559        It is guaranteed to each SCTP socket (but not association) even
1560        under moderate memory pressure.
1561
1562        Default: 1 page
1563
1564sctp_wmem  - vector of 3 INTEGERs: min, default, max
1565        Currently this tunable has no effect.
1566
1567addr_scope_policy - INTEGER
1568        Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1569
1570        0   - Disable IPv4 address scoping
1571        1   - Enable IPv4 address scoping
1572        2   - Follow draft but allow IPv4 private addresses
1573        3   - Follow draft but allow IPv4 link local addresses
1574
1575        Default: 1
1576
1577
1578/proc/sys/net/core/*
1579        Please see: Documentation/sysctl/net.txt for descriptions of these entries.
1580
1581
1582/proc/sys/net/unix/*
1583max_dgram_qlen - INTEGER
1584        The maximum length of dgram socket receive queue
1585
1586        Default: 10
1587
1588
1589UNDOCUMENTED:
1590
1591/proc/sys/net/irda/*
1592        fast_poll_increase FIXME
1593        warn_noreply_time FIXME
1594        discovery_slots FIXME
1595        slot_timeout FIXME
1596        max_baud_rate FIXME
1597        discovery_timeout FIXME
1598        lap_keepalive_time FIXME
1599        max_noreply_time FIXME
1600        max_tx_data_size FIXME
1601        max_tx_window FIXME
1602        min_tx_turn_time FIXME
1603
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.