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 - INTEGER
  19        Disable Path MTU Discovery. If enabled in mode 1 and a
  20        fragmentation-required ICMP is received, the PMTU to this
  21        destination will be set to min_pmtu (see below). You will need
  22        to raise min_pmtu to the smallest interface MTU on your system
  23        manually if you want to avoid locally generated fragments.
  24
  25        In mode 2 incoming Path MTU Discovery messages will be
  26        discarded. Outgoing frames are handled the same as in mode 1,
  27        implicitly setting IP_PMTUDISC_DONT on every created socket.
  28
  29        Mode 3 is a hardend pmtu discover mode. The kernel will only
  30        accept fragmentation-needed errors if the underlying protocol
  31        can verify them besides a plain socket lookup. Current
  32        protocols for which pmtu events will be honored are TCP, SCTP
  33        and DCCP as they verify e.g. the sequence number or the
  34        association. This mode should not be enabled globally but is
  35        only intended to secure e.g. name servers in namespaces where
  36        TCP path mtu must still work but path MTU information of other
  37        protocols should be discarded. If enabled globally this mode
  38        could break other protocols.
  39
  40        Possible values: 0-3
  41        Default: FALSE
  42
  43min_pmtu - INTEGER
  44        default 552 - minimum discovered Path MTU
  45
  46ip_forward_use_pmtu - BOOLEAN
  47        By default we don't trust protocol path MTUs while forwarding
  48        because they could be easily forged and can lead to unwanted
  49        fragmentation by the router.
  50        You only need to enable this if you have user-space software
  51        which tries to discover path mtus by itself and depends on the
  52        kernel honoring this information. This is normally not the
  53        case.
  54        Default: 0 (disabled)
  55        Possible values:
  56        0 - disabled
  57        1 - enabled
  58
  59route/max_size - INTEGER
  60        Maximum number of routes allowed in the kernel.  Increase
  61        this when using large numbers of interfaces and/or routes.
  62
  63neigh/default/gc_thresh1 - INTEGER
  64        Minimum number of entries to keep.  Garbage collector will not
  65        purge entries if there are fewer than this number.
  66        Default: 128
  67
  68neigh/default/gc_thresh3 - INTEGER
  69        Maximum number of neighbor entries allowed.  Increase this
  70        when using large numbers of interfaces and when communicating
  71        with large numbers of directly-connected peers.
  72        Default: 1024
  73
  74neigh/default/unres_qlen_bytes - INTEGER
  75        The maximum number of bytes which may be used by packets
  76        queued for each unresolved address by other network layers.
  77        (added in linux 3.3)
  78        Setting negative value is meaningless and will return error.
  79        Default: 65536 Bytes(64KB)
  80
  81neigh/default/unres_qlen - INTEGER
  82        The maximum number of packets which may be queued for each
  83        unresolved address by other network layers.
  84        (deprecated in linux 3.3) : use unres_qlen_bytes instead.
  85        Prior to linux 3.3, the default value is 3 which may cause
  86        unexpected packet loss. The current default value is calculated
  87        according to default value of unres_qlen_bytes and true size of
  88        packet.
  89        Default: 31
  90
  91mtu_expires - INTEGER
  92        Time, in seconds, that cached PMTU information is kept.
  93
  94min_adv_mss - INTEGER
  95        The advertised MSS depends on the first hop route MTU, but will
  96        never be lower than this setting.
  97
  98IP Fragmentation:
  99
 100ipfrag_high_thresh - INTEGER
 101        Maximum memory used to reassemble IP fragments. When
 102        ipfrag_high_thresh bytes of memory is allocated for this purpose,
 103        the fragment handler will toss packets until ipfrag_low_thresh
 104        is reached.
 105
 106ipfrag_low_thresh - INTEGER
 107        See ipfrag_high_thresh
 108
 109ipfrag_time - INTEGER
 110        Time in seconds to keep an IP fragment in memory.
 111
 112ipfrag_secret_interval - INTEGER
 113        Regeneration interval (in seconds) of the hash secret (or lifetime
 114        for the hash secret) for IP fragments.
 115        Default: 600
 116
 117ipfrag_max_dist - INTEGER
 118        ipfrag_max_dist is a non-negative integer value which defines the
 119        maximum "disorder" which is allowed among fragments which share a
 120        common IP source address. Note that reordering of packets is
 121        not unusual, but if a large number of fragments arrive from a source
 122        IP address while a particular fragment queue remains incomplete, it
 123        probably indicates that one or more fragments belonging to that queue
 124        have been lost. When ipfrag_max_dist is positive, an additional check
 125        is done on fragments before they are added to a reassembly queue - if
 126        ipfrag_max_dist (or more) fragments have arrived from a particular IP
 127        address between additions to any IP fragment queue using that source
 128        address, it's presumed that one or more fragments in the queue are
 129        lost. The existing fragment queue will be dropped, and a new one
 130        started. An ipfrag_max_dist value of zero disables this check.
 131
 132        Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
 133        result in unnecessarily dropping fragment queues when normal
 134        reordering of packets occurs, which could lead to poor application
 135        performance. Using a very large value, e.g. 50000, increases the
 136        likelihood of incorrectly reassembling IP fragments that originate
 137        from different IP datagrams, which could result in data corruption.
 138        Default: 64
 139
 140INET peer storage:
 141
 142inet_peer_threshold - INTEGER
 143        The approximate size of the storage.  Starting from this threshold
 144        entries will be thrown aggressively.  This threshold also determines
 145        entries' time-to-live and time intervals between garbage collection
 146        passes.  More entries, less time-to-live, less GC interval.
 147
 148inet_peer_minttl - INTEGER
 149        Minimum time-to-live of entries.  Should be enough to cover fragment
 150        time-to-live on the reassembling side.  This minimum time-to-live  is
 151        guaranteed if the pool size is less than inet_peer_threshold.
 152        Measured in seconds.
 153
 154inet_peer_maxttl - INTEGER
 155        Maximum time-to-live of entries.  Unused entries will expire after
 156        this period of time if there is no memory pressure on the pool (i.e.
 157        when the number of entries in the pool is very small).
 158        Measured in seconds.
 159
 160TCP variables:
 161
 162somaxconn - INTEGER
 163        Limit of socket listen() backlog, known in userspace as SOMAXCONN.
 164        Defaults to 128.  See also tcp_max_syn_backlog for additional tuning
 165        for TCP sockets.
 166
 167tcp_abort_on_overflow - BOOLEAN
 168        If listening service is too slow to accept new connections,
 169        reset them. Default state is FALSE. It means that if overflow
 170        occurred due to a burst, connection will recover. Enable this
 171        option _only_ if you are really sure that listening daemon
 172        cannot be tuned to accept connections faster. Enabling this
 173        option can harm clients of your server.
 174
 175tcp_adv_win_scale - INTEGER
 176        Count buffering overhead as bytes/2^tcp_adv_win_scale
 177        (if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
 178        if it is <= 0.
 179        Possible values are [-31, 31], inclusive.
 180        Default: 1
 181
 182tcp_allowed_congestion_control - STRING
 183        Show/set the congestion control choices available to non-privileged
 184        processes. The list is a subset of those listed in
 185        tcp_available_congestion_control.
 186        Default is "reno" and the default setting (tcp_congestion_control).
 187
 188tcp_app_win - INTEGER
 189        Reserve max(window/2^tcp_app_win, mss) of window for application
 190        buffer. Value 0 is special, it means that nothing is reserved.
 191        Default: 31
 192
 193tcp_autocorking - BOOLEAN
 194        Enable TCP auto corking :
 195        When applications do consecutive small write()/sendmsg() system calls,
 196        we try to coalesce these small writes as much as possible, to lower
 197        total amount of sent packets. This is done if at least one prior
 198        packet for the flow is waiting in Qdisc queues or device transmit
 199        queue. Applications can still use TCP_CORK for optimal behavior
 200        when they know how/when to uncork their sockets.
 201        Default : 1
 202
 203tcp_available_congestion_control - STRING
 204        Shows the available congestion control choices that are registered.
 205        More congestion control algorithms may be available as modules,
 206        but not loaded.
 207
 208tcp_base_mss - INTEGER
 209        The initial value of search_low to be used by the packetization layer
 210        Path MTU discovery (MTU probing).  If MTU probing is enabled,
 211        this is the initial MSS used by the connection.
 212
 213tcp_congestion_control - STRING
 214        Set the congestion control algorithm to be used for new
 215        connections. The algorithm "reno" is always available, but
 216        additional choices may be available based on kernel configuration.
 217        Default is set as part of kernel configuration.
 218        For passive connections, the listener congestion control choice
 219        is inherited.
 220        [see setsockopt(listenfd, SOL_TCP, TCP_CONGESTION, "name" ...) ]
 221
 222tcp_dsack - BOOLEAN
 223        Allows TCP to send "duplicate" SACKs.
 224
 225tcp_early_retrans - INTEGER
 226        Enable Early Retransmit (ER), per RFC 5827. ER lowers the threshold
 227        for triggering fast retransmit when the amount of outstanding data is
 228        small and when no previously unsent data can be transmitted (such
 229        that limited transmit could be used). Also controls the use of
 230        Tail loss probe (TLP) that converts RTOs occurring due to tail
 231        losses into fast recovery (draft-dukkipati-tcpm-tcp-loss-probe-01).
 232        Possible values:
 233                0 disables ER
 234                1 enables ER
 235                2 enables ER but delays fast recovery and fast retransmit
 236                  by a fourth of RTT. This mitigates connection falsely
 237                  recovers when network has a small degree of reordering
 238                  (less than 3 packets).
 239                3 enables delayed ER and TLP.
 240                4 enables TLP only.
 241        Default: 3
 242
 243tcp_ecn - INTEGER
 244        Control use of Explicit Congestion Notification (ECN) by TCP.
 245        ECN is used only when both ends of the TCP connection indicate
 246        support for it.  This feature is useful in avoiding losses due
 247        to congestion by allowing supporting routers to signal
 248        congestion before having to drop packets.
 249        Possible values are:
 250                0 Disable ECN.  Neither initiate nor accept ECN.
 251                1 Enable ECN when requested by incoming connections and
 252                  also request ECN on outgoing connection attempts.
 253                2 Enable ECN when requested by incoming connections
 254                  but do not request ECN on outgoing connections.
 255        Default: 2
 256
 257tcp_fack - BOOLEAN
 258        Enable FACK congestion avoidance and fast retransmission.
 259        The value is not used, if tcp_sack is not enabled.
 260
 261tcp_fin_timeout - INTEGER
 262        The length of time an orphaned (no longer referenced by any
 263        application) connection will remain in the FIN_WAIT_2 state
 264        before it is aborted at the local end.  While a perfectly
 265        valid "receive only" state for an un-orphaned connection, an
 266        orphaned connection in FIN_WAIT_2 state could otherwise wait
 267        forever for the remote to close its end of the connection.
 268        Cf. tcp_max_orphans
 269        Default: 60 seconds
 270
 271tcp_frto - INTEGER
 272        Enables Forward RTO-Recovery (F-RTO) defined in RFC5682.
 273        F-RTO is an enhanced recovery algorithm for TCP retransmission
 274        timeouts.  It is particularly beneficial in networks where the
 275        RTT fluctuates (e.g., wireless). F-RTO is sender-side only
 276        modification. It does not require any support from the peer.
 277
 278        By default it's enabled with a non-zero value. 0 disables F-RTO.
 279
 280tcp_keepalive_time - INTEGER
 281        How often TCP sends out keepalive messages when keepalive is enabled.
 282        Default: 2hours.
 283
 284tcp_keepalive_probes - INTEGER
 285        How many keepalive probes TCP sends out, until it decides that the
 286        connection is broken. Default value: 9.
 287
 288tcp_keepalive_intvl - INTEGER
 289        How frequently the probes are send out. Multiplied by
 290        tcp_keepalive_probes it is time to kill not responding connection,
 291        after probes started. Default value: 75sec i.e. connection
 292        will be aborted after ~11 minutes of retries.
 293
 294tcp_low_latency - BOOLEAN
 295        If set, the TCP stack makes decisions that prefer lower
 296        latency as opposed to higher throughput.  By default, this
 297        option is not set meaning that higher throughput is preferred.
 298        An example of an application where this default should be
 299        changed would be a Beowulf compute cluster.
 300        Default: 0
 301
 302tcp_max_orphans - INTEGER
 303        Maximal number of TCP sockets not attached to any user file handle,
 304        held by system. If this number is exceeded orphaned connections are
 305        reset immediately and warning is printed. This limit exists
 306        only to prevent simple DoS attacks, you _must_ not rely on this
 307        or lower the limit artificially, but rather increase it
 308        (probably, after increasing installed memory),
 309        if network conditions require more than default value,
 310        and tune network services to linger and kill such states
 311        more aggressively. Let me to remind again: each orphan eats
 312        up to ~64K of unswappable memory.
 313
 314tcp_max_syn_backlog - INTEGER
 315        Maximal number of remembered connection requests, which have not
 316        received an acknowledgment from connecting client.
 317        The minimal value is 128 for low memory machines, and it will
 318        increase in proportion to the memory of machine.
 319        If server suffers from overload, try increasing this number.
 320
 321tcp_max_tw_buckets - INTEGER
 322        Maximal number of timewait sockets held by system simultaneously.
 323        If this number is exceeded time-wait socket is immediately destroyed
 324        and warning is printed. This limit exists only to prevent
 325        simple DoS attacks, you _must_ not lower the limit artificially,
 326        but rather increase it (probably, after increasing installed memory),
 327        if network conditions require more than default value.
 328
 329tcp_mem - vector of 3 INTEGERs: min, pressure, max
 330        min: below this number of pages TCP is not bothered about its
 331        memory appetite.
 332
 333        pressure: when amount of memory allocated by TCP exceeds this number
 334        of pages, TCP moderates its memory consumption and enters memory
 335        pressure mode, which is exited when memory consumption falls
 336        under "min".
 337
 338        max: number of pages allowed for queueing by all TCP sockets.
 339
 340        Defaults are calculated at boot time from amount of available
 341        memory.
 342
 343tcp_moderate_rcvbuf - BOOLEAN
 344        If set, TCP performs receive buffer auto-tuning, attempting to
 345        automatically size the buffer (no greater than tcp_rmem[2]) to
 346        match the size required by the path for full throughput.  Enabled by
 347        default.
 348
 349tcp_mtu_probing - INTEGER
 350        Controls TCP Packetization-Layer Path MTU Discovery.  Takes three
 351        values:
 352          0 - Disabled
 353          1 - Disabled by default, enabled when an ICMP black hole detected
 354          2 - Always enabled, use initial MSS of tcp_base_mss.
 355
 356tcp_no_metrics_save - BOOLEAN
 357        By default, TCP saves various connection metrics in the route cache
 358        when the connection closes, so that connections established in the
 359        near future can use these to set initial conditions.  Usually, this
 360        increases overall performance, but may sometimes cause performance
 361        degradation.  If set, TCP will not cache metrics on closing
 362        connections.
 363
 364tcp_orphan_retries - INTEGER
 365        This value influences the timeout of a locally closed TCP connection,
 366        when RTO retransmissions remain unacknowledged.
 367        See tcp_retries2 for more details.
 368
 369        The default value is 8.
 370        If your machine is a loaded WEB server,
 371        you should think about lowering this value, such sockets
 372        may consume significant resources. Cf. tcp_max_orphans.
 373
 374tcp_reordering - INTEGER
 375        Maximal reordering of packets in a TCP stream.
 376        Default: 3
 377
 378tcp_retrans_collapse - BOOLEAN
 379        Bug-to-bug compatibility with some broken printers.
 380        On retransmit try to send bigger packets to work around bugs in
 381        certain TCP stacks.
 382
 383tcp_retries1 - INTEGER
 384        This value influences the time, after which TCP decides, that
 385        something is wrong due to unacknowledged RTO retransmissions,
 386        and reports this suspicion to the network layer.
 387        See tcp_retries2 for more details.
 388
 389        RFC 1122 recommends at least 3 retransmissions, which is the
 390        default.
 391
 392tcp_retries2 - INTEGER
 393        This value influences the timeout of an alive TCP connection,
 394        when RTO retransmissions remain unacknowledged.
 395        Given a value of N, a hypothetical TCP connection following
 396        exponential backoff with an initial RTO of TCP_RTO_MIN would
 397        retransmit N times before killing the connection at the (N+1)th RTO.
 398
 399        The default value of 15 yields a hypothetical timeout of 924.6
 400        seconds and is a lower bound for the effective timeout.
 401        TCP will effectively time out at the first RTO which exceeds the
 402        hypothetical timeout.
 403
 404        RFC 1122 recommends at least 100 seconds for the timeout,
 405        which corresponds to a value of at least 8.
 406
 407tcp_rfc1337 - BOOLEAN
 408        If set, the TCP stack behaves conforming to RFC1337. If unset,
 409        we are not conforming to RFC, but prevent TCP TIME_WAIT
 410        assassination.
 411        Default: 0
 412
 413tcp_rmem - vector of 3 INTEGERs: min, default, max
 414        min: Minimal size of receive buffer used by TCP sockets.
 415        It is guaranteed to each TCP socket, even under moderate memory
 416        pressure.
 417        Default: 1 page
 418
 419        default: initial size of receive buffer used by TCP sockets.
 420        This value overrides net.core.rmem_default used by other protocols.
 421        Default: 87380 bytes. This value results in window of 65535 with
 422        default setting of tcp_adv_win_scale and tcp_app_win:0 and a bit
 423        less for default tcp_app_win. See below about these variables.
 424
 425        max: maximal size of receive buffer allowed for automatically
 426        selected receiver buffers for TCP socket. This value does not override
 427        net.core.rmem_max.  Calling setsockopt() with SO_RCVBUF disables
 428        automatic tuning of that socket's receive buffer size, in which
 429        case this value is ignored.
 430        Default: between 87380B and 6MB, depending on RAM size.
 431
 432tcp_sack - BOOLEAN
 433        Enable select acknowledgments (SACKS).
 434
 435tcp_slow_start_after_idle - BOOLEAN
 436        If set, provide RFC2861 behavior and time out the congestion
 437        window after an idle period.  An idle period is defined at
 438        the current RTO.  If unset, the congestion window will not
 439        be timed out after an idle period.
 440        Default: 1
 441
 442tcp_stdurg - BOOLEAN
 443        Use the Host requirements interpretation of the TCP urgent pointer field.
 444        Most hosts use the older BSD interpretation, so if you turn this on
 445        Linux might not communicate correctly with them.
 446        Default: FALSE
 447
 448tcp_synack_retries - INTEGER
 449        Number of times SYNACKs for a passive TCP connection attempt will
 450        be retransmitted. Should not be higher than 255. Default value
 451        is 5, which corresponds to 31seconds till the last retransmission
 452        with the current initial RTO of 1second. With this the final timeout
 453        for a passive TCP connection will happen after 63seconds.
 454
 455tcp_syncookies - BOOLEAN
 456        Only valid when the kernel was compiled with CONFIG_SYN_COOKIES
 457        Send out syncookies when the syn backlog queue of a socket
 458        overflows. This is to prevent against the common 'SYN flood attack'
 459        Default: 1
 460
 461        Note, that syncookies is fallback facility.
 462        It MUST NOT be used to help highly loaded servers to stand
 463        against legal connection rate. If you see SYN flood warnings
 464        in your logs, but investigation shows that they occur
 465        because of overload with legal connections, you should tune
 466        another parameters until this warning disappear.
 467        See: tcp_max_syn_backlog, tcp_synack_retries, tcp_abort_on_overflow.
 468
 469        syncookies seriously violate TCP protocol, do not allow
 470        to use TCP extensions, can result in serious degradation
 471        of some services (f.e. SMTP relaying), visible not by you,
 472        but your clients and relays, contacting you. While you see
 473        SYN flood warnings in logs not being really flooded, your server
 474        is seriously misconfigured.
 475
 476        If you want to test which effects syncookies have to your
 477        network connections you can set this knob to 2 to enable
 478        unconditionally generation of syncookies.
 479
 480tcp_fastopen - INTEGER
 481        Enable TCP Fast Open feature (draft-ietf-tcpm-fastopen) to send data
 482        in the opening SYN packet. To use this feature, the client application
 483        must use sendmsg() or sendto() with MSG_FASTOPEN flag rather than
 484        connect() to perform a TCP handshake automatically.
 485
 486        The values (bitmap) are
 487        1: Enables sending data in the opening SYN on the client w/ MSG_FASTOPEN.
 488        2: Enables TCP Fast Open on the server side, i.e., allowing data in
 489           a SYN packet to be accepted and passed to the application before
 490           3-way hand shake finishes.
 491        4: Send data in the opening SYN regardless of cookie availability and
 492           without a cookie option.
 493        0x100: Accept SYN data w/o validating the cookie.
 494        0x200: Accept data-in-SYN w/o any cookie option present.
 495        0x400/0x800: Enable Fast Open on all listeners regardless of the
 496           TCP_FASTOPEN socket option. The two different flags designate two
 497           different ways of setting max_qlen without the TCP_FASTOPEN socket
 498           option.
 499
 500        Default: 1
 501
 502        Note that the client & server side Fast Open flags (1 and 2
 503        respectively) must be also enabled before the rest of flags can take
 504        effect.
 505
 506        See include/net/tcp.h and the code for more details.
 507
 508tcp_syn_retries - INTEGER
 509        Number of times initial SYNs for an active TCP connection attempt
 510        will be retransmitted. Should not be higher than 255. Default value
 511        is 6, which corresponds to 63seconds till the last retransmission
 512        with the current initial RTO of 1second. With this the final timeout
 513        for an active TCP connection attempt will happen after 127seconds.
 514
 515tcp_timestamps - BOOLEAN
 516        Enable timestamps as defined in RFC1323.
 517
 518tcp_min_tso_segs - INTEGER
 519        Minimal number of segments per TSO frame.
 520        Since linux-3.12, TCP does an automatic sizing of TSO frames,
 521        depending on flow rate, instead of filling 64Kbytes packets.
 522        For specific usages, it's possible to force TCP to build big
 523        TSO frames. Note that TCP stack might split too big TSO packets
 524        if available window is too small.
 525        Default: 2
 526
 527tcp_tso_win_divisor - INTEGER
 528        This allows control over what percentage of the congestion window
 529        can be consumed by a single TSO frame.
 530        The setting of this parameter is a choice between burstiness and
 531        building larger TSO frames.
 532        Default: 3
 533
 534tcp_tw_recycle - BOOLEAN
 535        Enable fast recycling TIME-WAIT sockets. Default value is 0.
 536        It should not be changed without advice/request of technical
 537        experts.
 538
 539tcp_tw_reuse - BOOLEAN
 540        Allow to reuse TIME-WAIT sockets for new connections when it is
 541        safe from protocol viewpoint. Default value is 0.
 542        It should not be changed without advice/request of technical
 543        experts.
 544
 545tcp_window_scaling - BOOLEAN
 546        Enable window scaling as defined in RFC1323.
 547
 548tcp_wmem - vector of 3 INTEGERs: min, default, max
 549        min: Amount of memory reserved for send buffers for TCP sockets.
 550        Each TCP socket has rights to use it due to fact of its birth.
 551        Default: 1 page
 552
 553        default: initial size of send buffer used by TCP sockets.  This
 554        value overrides net.core.wmem_default used by other protocols.
 555        It is usually lower than net.core.wmem_default.
 556        Default: 16K
 557
 558        max: Maximal amount of memory allowed for automatically tuned
 559        send buffers for TCP sockets. This value does not override
 560        net.core.wmem_max.  Calling setsockopt() with SO_SNDBUF disables
 561        automatic tuning of that socket's send buffer size, in which case
 562        this value is ignored.
 563        Default: between 64K and 4MB, depending on RAM size.
 564
 565tcp_notsent_lowat - UNSIGNED INTEGER
 566        A TCP socket can control the amount of unsent bytes in its write queue,
 567        thanks to TCP_NOTSENT_LOWAT socket option. poll()/select()/epoll()
 568        reports POLLOUT events if the amount of unsent bytes is below a per
 569        socket value, and if the write queue is not full. sendmsg() will
 570        also not add new buffers if the limit is hit.
 571
 572        This global variable controls the amount of unsent data for
 573        sockets not using TCP_NOTSENT_LOWAT. For these sockets, a change
 574        to the global variable has immediate effect.
 575
 576        Default: UINT_MAX (0xFFFFFFFF)
 577
 578tcp_workaround_signed_windows - BOOLEAN
 579        If set, assume no receipt of a window scaling option means the
 580        remote TCP is broken and treats the window as a signed quantity.
 581        If unset, assume the remote TCP is not broken even if we do
 582        not receive a window scaling option from them.
 583        Default: 0
 584
 585tcp_dma_copybreak - INTEGER
 586        Lower limit, in bytes, of the size of socket reads that will be
 587        offloaded to a DMA copy engine, if one is present in the system
 588        and CONFIG_NET_DMA is enabled.
 589        Default: 4096
 590
 591tcp_thin_linear_timeouts - BOOLEAN
 592        Enable dynamic triggering of linear timeouts for thin streams.
 593        If set, a check is performed upon retransmission by timeout to
 594        determine if the stream is thin (less than 4 packets in flight).
 595        As long as the stream is found to be thin, up to 6 linear
 596        timeouts may be performed before exponential backoff mode is
 597        initiated. This improves retransmission latency for
 598        non-aggressive thin streams, often found to be time-dependent.
 599        For more information on thin streams, see
 600        Documentation/networking/tcp-thin.txt
 601        Default: 0
 602
 603tcp_thin_dupack - BOOLEAN
 604        Enable dynamic triggering of retransmissions after one dupACK
 605        for thin streams. If set, a check is performed upon reception
 606        of a dupACK to determine if the stream is thin (less than 4
 607        packets in flight). As long as the stream is found to be thin,
 608        data is retransmitted on the first received dupACK. This
 609        improves retransmission latency for non-aggressive thin
 610        streams, often found to be time-dependent.
 611        For more information on thin streams, see
 612        Documentation/networking/tcp-thin.txt
 613        Default: 0
 614
 615tcp_limit_output_bytes - INTEGER
 616        Controls TCP Small Queue limit per tcp socket.
 617        TCP bulk sender tends to increase packets in flight until it
 618        gets losses notifications. With SNDBUF autotuning, this can
 619        result in a large amount of packets queued in qdisc/device
 620        on the local machine, hurting latency of other flows, for
 621        typical pfifo_fast qdiscs.
 622        tcp_limit_output_bytes limits the number of bytes on qdisc
 623        or device to reduce artificial RTT/cwnd and reduce bufferbloat.
 624        Default: 131072
 625
 626tcp_challenge_ack_limit - INTEGER
 627        Limits number of Challenge ACK sent per second, as recommended
 628        in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
 629        Default: 100
 630
 631UDP variables:
 632
 633udp_mem - vector of 3 INTEGERs: min, pressure, max
 634        Number of pages allowed for queueing by all UDP sockets.
 635
 636        min: Below this number of pages UDP is not bothered about its
 637        memory appetite. When amount of memory allocated by UDP exceeds
 638        this number, UDP starts to moderate memory usage.
 639
 640        pressure: This value was introduced to follow format of tcp_mem.
 641
 642        max: Number of pages allowed for queueing by all UDP sockets.
 643
 644        Default is calculated at boot time from amount of available memory.
 645
 646udp_rmem_min - INTEGER
 647        Minimal size of receive buffer used by UDP sockets in moderation.
 648        Each UDP socket is able to use the size for receiving data, even if
 649        total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
 650        Default: 1 page
 651
 652udp_wmem_min - INTEGER
 653        Minimal size of send buffer used by UDP sockets in moderation.
 654        Each UDP socket is able to use the size for sending data, even if
 655        total pages of UDP sockets exceed udp_mem pressure. The unit is byte.
 656        Default: 1 page
 657
 658CIPSOv4 Variables:
 659
 660cipso_cache_enable - BOOLEAN
 661        If set, enable additions to and lookups from the CIPSO label mapping
 662        cache.  If unset, additions are ignored and lookups always result in a
 663        miss.  However, regardless of the setting the cache is still
 664        invalidated when required when means you can safely toggle this on and
 665        off and the cache will always be "safe".
 666        Default: 1
 667
 668cipso_cache_bucket_size - INTEGER
 669        The CIPSO label cache consists of a fixed size hash table with each
 670        hash bucket containing a number of cache entries.  This variable limits
 671        the number of entries in each hash bucket; the larger the value the
 672        more CIPSO label mappings that can be cached.  When the number of
 673        entries in a given hash bucket reaches this limit adding new entries
 674        causes the oldest entry in the bucket to be removed to make room.
 675        Default: 10
 676
 677cipso_rbm_optfmt - BOOLEAN
 678        Enable the "Optimized Tag 1 Format" as defined in section 3.4.2.6 of
 679        the CIPSO draft specification (see Documentation/netlabel for details).
 680        This means that when set the CIPSO tag will be padded with empty
 681        categories in order to make the packet data 32-bit aligned.
 682        Default: 0
 683
 684cipso_rbm_structvalid - BOOLEAN
 685        If set, do a very strict check of the CIPSO option when
 686        ip_options_compile() is called.  If unset, relax the checks done during
 687        ip_options_compile().  Either way is "safe" as errors are caught else
 688        where in the CIPSO processing code but setting this to 0 (False) should
 689        result in less work (i.e. it should be faster) but could cause problems
 690        with other implementations that require strict checking.
 691        Default: 0
 692
 693IP Variables:
 694
 695ip_local_port_range - 2 INTEGERS
 696        Defines the local port range that is used by TCP and UDP to
 697        choose the local port. The first number is the first, the
 698        second the last local port number. The default values are
 699        32768 and 61000 respectively.
 700
 701ip_local_reserved_ports - list of comma separated ranges
 702        Specify the ports which are reserved for known third-party
 703        applications. These ports will not be used by automatic port
 704        assignments (e.g. when calling connect() or bind() with port
 705        number 0). Explicit port allocation behavior is unchanged.
 706
 707        The format used for both input and output is a comma separated
 708        list of ranges (e.g. "1,2-4,10-10" for ports 1, 2, 3, 4 and
 709        10). Writing to the file will clear all previously reserved
 710        ports and update the current list with the one given in the
 711        input.
 712
 713        Note that ip_local_port_range and ip_local_reserved_ports
 714        settings are independent and both are considered by the kernel
 715        when determining which ports are available for automatic port
 716        assignments.
 717
 718        You can reserve ports which are not in the current
 719        ip_local_port_range, e.g.:
 720
 721        $ cat /proc/sys/net/ipv4/ip_local_port_range
 722        32000   61000
 723        $ cat /proc/sys/net/ipv4/ip_local_reserved_ports
 724        8080,9148
 725
 726        although this is redundant. However such a setting is useful
 727        if later the port range is changed to a value that will
 728        include the reserved ports.
 729
 730        Default: Empty
 731
 732ip_nonlocal_bind - BOOLEAN
 733        If set, allows processes to bind() to non-local IP addresses,
 734        which can be quite useful - but may break some applications.
 735        Default: 0
 736
 737ip_dynaddr - BOOLEAN
 738        If set non-zero, enables support for dynamic addresses.
 739        If set to a non-zero value larger than 1, a kernel log
 740        message will be printed when dynamic address rewriting
 741        occurs.
 742        Default: 0
 743
 744ip_early_demux - BOOLEAN
 745        Optimize input packet processing down to one demux for
 746        certain kinds of local sockets.  Currently we only do this
 747        for established TCP sockets.
 748
 749        It may add an additional cost for pure routing workloads that
 750        reduces overall throughput, in such case you should disable it.
 751        Default: 1
 752
 753icmp_echo_ignore_all - BOOLEAN
 754        If set non-zero, then the kernel will ignore all ICMP ECHO
 755        requests sent to it.
 756        Default: 0
 757
 758icmp_echo_ignore_broadcasts - BOOLEAN
 759        If set non-zero, then the kernel will ignore all ICMP ECHO and
 760        TIMESTAMP requests sent to it via broadcast/multicast.
 761        Default: 1
 762
 763icmp_ratelimit - INTEGER
 764        Limit the maximal rates for sending ICMP packets whose type matches
 765        icmp_ratemask (see below) to specific targets.
 766        0 to disable any limiting,
 767        otherwise the minimal space between responses in milliseconds.
 768        Default: 1000
 769
 770icmp_ratemask - INTEGER
 771        Mask made of ICMP types for which rates are being limited.
 772        Significant bits: IHGFEDCBA9876543210
 773        Default mask:     0000001100000011000 (6168)
 774
 775        Bit definitions (see include/linux/icmp.h):
 776                0 Echo Reply
 777                3 Destination Unreachable *
 778                4 Source Quench *
 779                5 Redirect
 780                8 Echo Request
 781                B Time Exceeded *
 782                C Parameter Problem *
 783                D Timestamp Request
 784                E Timestamp Reply
 785                F Info Request
 786                G Info Reply
 787                H Address Mask Request
 788                I Address Mask Reply
 789
 790        * These are rate limited by default (see default mask above)
 791
 792icmp_ignore_bogus_error_responses - BOOLEAN
 793        Some routers violate RFC1122 by sending bogus responses to broadcast
 794        frames.  Such violations are normally logged via a kernel warning.
 795        If this is set to TRUE, the kernel will not give such warnings, which
 796        will avoid log file clutter.
 797        Default: 1
 798
 799icmp_errors_use_inbound_ifaddr - BOOLEAN
 800
 801        If zero, icmp error messages are sent with the primary address of
 802        the exiting interface.
 803
 804        If non-zero, the message will be sent with the primary address of
 805        the interface that received the packet that caused the icmp error.
 806        This is the behaviour network many administrators will expect from
 807        a router. And it can make debugging complicated network layouts
 808        much easier.
 809
 810        Note that if no primary address exists for the interface selected,
 811        then the primary address of the first non-loopback interface that
 812        has one will be used regardless of this setting.
 813
 814        Default: 0
 815
 816igmp_max_memberships - INTEGER
 817        Change the maximum number of multicast groups we can subscribe to.
 818        Default: 20
 819
 820        Theoretical maximum value is bounded by having to send a membership
 821        report in a single datagram (i.e. the report can't span multiple
 822        datagrams, or risk confusing the switch and leaving groups you don't
 823        intend to).
 824
 825        The number of supported groups 'M' is bounded by the number of group
 826        report entries you can fit into a single datagram of 65535 bytes.
 827
 828        M = 65536-sizeof (ip header)/(sizeof(Group record))
 829
 830        Group records are variable length, with a minimum of 12 bytes.
 831        So net.ipv4.igmp_max_memberships should not be set higher than:
 832
 833        (65536-24) / 12 = 5459
 834
 835        The value 5459 assumes no IP header options, so in practice
 836        this number may be lower.
 837
 838        conf/interface/*  changes special settings per interface (where
 839        "interface" is the name of your network interface)
 840
 841        conf/all/*        is special, changes the settings for all interfaces
 842
 843log_martians - BOOLEAN
 844        Log packets with impossible addresses to kernel log.
 845        log_martians for the interface will be enabled if at least one of
 846        conf/{all,interface}/log_martians is set to TRUE,
 847        it will be disabled otherwise
 848
 849accept_redirects - BOOLEAN
 850        Accept ICMP redirect messages.
 851        accept_redirects for the interface will be enabled if:
 852        - both conf/{all,interface}/accept_redirects are TRUE in the case
 853          forwarding for the interface is enabled
 854        or
 855        - at least one of conf/{all,interface}/accept_redirects is TRUE in the
 856          case forwarding for the interface is disabled
 857        accept_redirects for the interface will be disabled otherwise
 858        default TRUE (host)
 859                FALSE (router)
 860
 861forwarding - BOOLEAN
 862        Enable IP forwarding on this interface.
 863
 864mc_forwarding - BOOLEAN
 865        Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE
 866        and a multicast routing daemon is required.
 867        conf/all/mc_forwarding must also be set to TRUE to enable multicast
 868        routing for the interface
 869
 870medium_id - INTEGER
 871        Integer value used to differentiate the devices by the medium they
 872        are attached to. Two devices can have different id values when
 873        the broadcast packets are received only on one of them.
 874        The default value 0 means that the device is the only interface
 875        to its medium, value of -1 means that medium is not known.
 876
 877        Currently, it is used to change the proxy_arp behavior:
 878        the proxy_arp feature is enabled for packets forwarded between
 879        two devices attached to different media.
 880
 881proxy_arp - BOOLEAN
 882        Do proxy arp.
 883        proxy_arp for the interface will be enabled if at least one of
 884        conf/{all,interface}/proxy_arp is set to TRUE,
 885        it will be disabled otherwise
 886
 887proxy_arp_pvlan - BOOLEAN
 888        Private VLAN proxy arp.
 889        Basically allow proxy arp replies back to the same interface
 890        (from which the ARP request/solicitation was received).
 891
 892        This is done to support (ethernet) switch features, like RFC
 893        3069, where the individual ports are NOT allowed to
 894        communicate with each other, but they are allowed to talk to
 895        the upstream router.  As described in RFC 3069, it is possible
 896        to allow these hosts to communicate through the upstream
 897        router by proxy_arp'ing. Don't need to be used together with
 898        proxy_arp.
 899
 900        This technology is known by different names:
 901          In RFC 3069 it is called VLAN Aggregation.
 902          Cisco and Allied Telesyn call it Private VLAN.
 903          Hewlett-Packard call it Source-Port filtering or port-isolation.
 904          Ericsson call it MAC-Forced Forwarding (RFC Draft).
 905
 906shared_media - BOOLEAN
 907        Send(router) or accept(host) RFC1620 shared media redirects.
 908        Overrides ip_secure_redirects.
 909        shared_media for the interface will be enabled if at least one of
 910        conf/{all,interface}/shared_media is set to TRUE,
 911        it will be disabled otherwise
 912        default TRUE
 913
 914secure_redirects - BOOLEAN
 915        Accept ICMP redirect messages only for gateways,
 916        listed in default gateway list.
 917        secure_redirects for the interface will be enabled if at least one of
 918        conf/{all,interface}/secure_redirects is set to TRUE,
 919        it will be disabled otherwise
 920        default TRUE
 921
 922send_redirects - BOOLEAN
 923        Send redirects, if router.
 924        send_redirects for the interface will be enabled if at least one of
 925        conf/{all,interface}/send_redirects is set to TRUE,
 926        it will be disabled otherwise
 927        Default: TRUE
 928
 929bootp_relay - BOOLEAN
 930        Accept packets with source address 0.b.c.d destined
 931        not to this host as local ones. It is supposed, that
 932        BOOTP relay daemon will catch and forward such packets.
 933        conf/all/bootp_relay must also be set to TRUE to enable BOOTP relay
 934        for the interface
 935        default FALSE
 936        Not Implemented Yet.
 937
 938accept_source_route - BOOLEAN
 939        Accept packets with SRR option.
 940        conf/all/accept_source_route must also be set to TRUE to accept packets
 941        with SRR option on the interface
 942        default TRUE (router)
 943                FALSE (host)
 944
 945accept_local - BOOLEAN
 946        Accept packets with local source addresses. In combination
 947        with suitable routing, this can be used to direct packets
 948        between two local interfaces over the wire and have them
 949        accepted properly.
 950
 951        rp_filter must be set to a non-zero value in order for
 952        accept_local to have an effect.
 953
 954        default FALSE
 955
 956route_localnet - BOOLEAN
 957        Do not consider loopback addresses as martian source or destination
 958        while routing. This enables the use of 127/8 for local routing purposes.
 959        default FALSE
 960
 961rp_filter - INTEGER
 962        0 - No source validation.
 963        1 - Strict mode as defined in RFC3704 Strict Reverse Path
 964            Each incoming packet is tested against the FIB and if the interface
 965            is not the best reverse path the packet check will fail.
 966            By default failed packets are discarded.
 967        2 - Loose mode as defined in RFC3704 Loose Reverse Path
 968            Each incoming packet's source address is also tested against the FIB
 969            and if the source address is not reachable via any interface
 970            the packet check will fail.
 971
 972        Current recommended practice in RFC3704 is to enable strict mode
 973        to prevent IP spoofing from DDos attacks. If using asymmetric routing
 974        or other complicated routing, then loose mode is recommended.
 975
 976        The max value from conf/{all,interface}/rp_filter is used
 977        when doing source validation on the {interface}.
 978
 979        Default value is 0. Note that some distributions enable it
 980        in startup scripts.
 981
 982arp_filter - BOOLEAN
 983        1 - Allows you to have multiple network interfaces on the same
 984        subnet, and have the ARPs for each interface be answered
 985        based on whether or not the kernel would route a packet from
 986        the ARP'd IP out that interface (therefore you must use source
 987        based routing for this to work). In other words it allows control
 988        of which cards (usually 1) will respond to an arp request.
 989
 990        0 - (default) The kernel can respond to arp requests with addresses
 991        from other interfaces. This may seem wrong but it usually makes
 992        sense, because it increases the chance of successful communication.
 993        IP addresses are owned by the complete host on Linux, not by
 994        particular interfaces. Only for more complex setups like load-
 995        balancing, does this behaviour cause problems.
 996
 997        arp_filter for the interface will be enabled if at least one of
 998        conf/{all,interface}/arp_filter is set to TRUE,
 999        it will be disabled otherwise
1000
1001arp_announce - INTEGER
1002        Define different restriction levels for announcing the local
1003        source IP address from IP packets in ARP requests sent on
1004        interface:
1005        0 - (default) Use any local address, configured on any interface
1006        1 - Try to avoid local addresses that are not in the target's
1007        subnet for this interface. This mode is useful when target
1008        hosts reachable via this interface require the source IP
1009        address in ARP requests to be part of their logical network
1010        configured on the receiving interface. When we generate the
1011        request we will check all our subnets that include the
1012        target IP and will preserve the source address if it is from
1013        such subnet. If there is no such subnet we select source
1014        address according to the rules for level 2.
1015        2 - Always use the best local address for this target.
1016        In this mode we ignore the source address in the IP packet
1017        and try to select local address that we prefer for talks with
1018        the target host. Such local address is selected by looking
1019        for primary IP addresses on all our subnets on the outgoing
1020        interface that include the target IP address. If no suitable
1021        local address is found we select the first local address
1022        we have on the outgoing interface or on all other interfaces,
1023        with the hope we will receive reply for our request and
1024        even sometimes no matter the source IP address we announce.
1025
1026        The max value from conf/{all,interface}/arp_announce is used.
1027
1028        Increasing the restriction level gives more chance for
1029        receiving answer from the resolved target while decreasing
1030        the level announces more valid sender's information.
1031
1032arp_ignore - INTEGER
1033        Define different modes for sending replies in response to
1034        received ARP requests that resolve local target IP addresses:
1035        0 - (default): reply for any local target IP address, configured
1036        on any interface
1037        1 - reply only if the target IP address is local address
1038        configured on the incoming interface
1039        2 - reply only if the target IP address is local address
1040        configured on the incoming interface and both with the
1041        sender's IP address are part from same subnet on this interface
1042        3 - do not reply for local addresses configured with scope host,
1043        only resolutions for global and link addresses are replied
1044        4-7 - reserved
1045        8 - do not reply for all local addresses
1046
1047        The max value from conf/{all,interface}/arp_ignore is used
1048        when ARP request is received on the {interface}
1049
1050arp_notify - BOOLEAN
1051        Define mode for notification of address and device changes.
1052        0 - (default): do nothing
1053        1 - Generate gratuitous arp requests when device is brought up
1054            or hardware address changes.
1055
1056arp_accept - BOOLEAN
1057        Define behavior for gratuitous ARP frames who's IP is not
1058        already present in the ARP table:
1059        0 - don't create new entries in the ARP table
1060        1 - create new entries in the ARP table
1061
1062        Both replies and requests type gratuitous arp will trigger the
1063        ARP table to be updated, if this setting is on.
1064
1065        If the ARP table already contains the IP address of the
1066        gratuitous arp frame, the arp table will be updated regardless
1067        if this setting is on or off.
1068
1069
1070app_solicit - INTEGER
1071        The maximum number of probes to send to the user space ARP daemon
1072        via netlink before dropping back to multicast probes (see
1073        mcast_solicit).  Defaults to 0.
1074
1075disable_policy - BOOLEAN
1076        Disable IPSEC policy (SPD) for this interface
1077
1078disable_xfrm - BOOLEAN
1079        Disable IPSEC encryption on this interface, whatever the policy
1080
1081igmpv2_unsolicited_report_interval - INTEGER
1082        The interval in milliseconds in which the next unsolicited
1083        IGMPv1 or IGMPv2 report retransmit will take place.
1084        Default: 10000 (10 seconds)
1085
1086igmpv3_unsolicited_report_interval - INTEGER
1087        The interval in milliseconds in which the next unsolicited
1088        IGMPv3 report retransmit will take place.
1089        Default: 1000 (1 seconds)
1090
1091promote_secondaries - BOOLEAN
1092        When a primary IP address is removed from this interface
1093        promote a corresponding secondary IP address instead of
1094        removing all the corresponding secondary IP addresses.
1095
1096
1097tag - INTEGER
1098        Allows you to write a number, which can be used as required.
1099        Default value is 0.
1100
1101Alexey Kuznetsov.
1102kuznet@ms2.inr.ac.ru
1103
1104Updated by:
1105Andi Kleen
1106ak@muc.de
1107Nicolas Delon
1108delon.nicolas@wanadoo.fr
1109
1110
1111
1112
1113/proc/sys/net/ipv6/* Variables:
1114
1115IPv6 has no global variables such as tcp_*.  tcp_* settings under ipv4/ also
1116apply to IPv6 [XXX?].
1117
1118bindv6only - BOOLEAN
1119        Default value for IPV6_V6ONLY socket option,
1120        which restricts use of the IPv6 socket to IPv6 communication
1121        only.
1122                TRUE: disable IPv4-mapped address feature
1123                FALSE: enable IPv4-mapped address feature
1124
1125        Default: FALSE (as specified in RFC3493)
1126
1127flowlabel_consistency - BOOLEAN
1128        Protect the consistency (and unicity) of flow label.
1129        You have to disable it to use IPV6_FL_F_REFLECT flag on the
1130        flow label manager.
1131        TRUE: enabled
1132        FALSE: disabled
1133        Default: TRUE
1134
1135anycast_src_echo_reply - BOOLEAN
1136        Controls the use of anycast addresses as source addresses for ICMPv6
1137        echo reply
1138        TRUE:  enabled
1139        FALSE: disabled
1140        Default: FALSE
1141
1142IPv6 Fragmentation:
1143
1144ip6frag_high_thresh - INTEGER
1145        Maximum memory used to reassemble IPv6 fragments. When
1146        ip6frag_high_thresh bytes of memory is allocated for this purpose,
1147        the fragment handler will toss packets until ip6frag_low_thresh
1148        is reached.
1149
1150ip6frag_low_thresh - INTEGER
1151        See ip6frag_high_thresh
1152
1153ip6frag_time - INTEGER
1154        Time in seconds to keep an IPv6 fragment in memory.
1155
1156ip6frag_secret_interval - INTEGER
1157        Regeneration interval (in seconds) of the hash secret (or lifetime
1158        for the hash secret) for IPv6 fragments.
1159        Default: 600
1160
1161conf/default/*:
1162        Change the interface-specific default settings.
1163
1164
1165conf/all/*:
1166        Change all the interface-specific settings.
1167
1168        [XXX:  Other special features than forwarding?]
1169
1170conf/all/forwarding - BOOLEAN
1171        Enable global IPv6 forwarding between all interfaces.
1172
1173        IPv4 and IPv6 work differently here; e.g. netfilter must be used
1174        to control which interfaces may forward packets and which not.
1175
1176        This also sets all interfaces' Host/Router setting
1177        'forwarding' to the specified value.  See below for details.
1178
1179        This referred to as global forwarding.
1180
1181proxy_ndp - BOOLEAN
1182        Do proxy ndp.
1183
1184conf/interface/*:
1185        Change special settings per interface.
1186
1187        The functional behaviour for certain settings is different
1188        depending on whether local forwarding is enabled or not.
1189
1190accept_ra - INTEGER
1191        Accept Router Advertisements; autoconfigure using them.
1192
1193        It also determines whether or not to transmit Router
1194        Solicitations. If and only if the functional setting is to
1195        accept Router Advertisements, Router Solicitations will be
1196        transmitted.
1197
1198        Possible values are:
1199                0 Do not accept Router Advertisements.
1200                1 Accept Router Advertisements if forwarding is disabled.
1201                2 Overrule forwarding behaviour. Accept Router Advertisements
1202                  even if forwarding is enabled.
1203
1204        Functional default: enabled if local forwarding is disabled.
1205                            disabled if local forwarding is enabled.
1206
1207accept_ra_defrtr - BOOLEAN
1208        Learn default router in Router Advertisement.
1209
1210        Functional default: enabled if accept_ra is enabled.
1211                            disabled if accept_ra is disabled.
1212
1213accept_ra_pinfo - BOOLEAN
1214        Learn Prefix Information in Router Advertisement.
1215
1216        Functional default: enabled if accept_ra is enabled.
1217                            disabled if accept_ra is disabled.
1218
1219accept_ra_rt_info_max_plen - INTEGER
1220        Maximum prefix length of Route Information in RA.
1221
1222        Route Information w/ prefix larger than or equal to this
1223        variable shall be ignored.
1224
1225        Functional default: 0 if accept_ra_rtr_pref is enabled.
1226                            -1 if accept_ra_rtr_pref is disabled.
1227
1228accept_ra_rtr_pref - BOOLEAN
1229        Accept Router Preference in RA.
1230
1231        Functional default: enabled if accept_ra is enabled.
1232                            disabled if accept_ra is disabled.
1233
1234accept_redirects - BOOLEAN
1235        Accept Redirects.
1236
1237        Functional default: enabled if local forwarding is disabled.
1238                            disabled if local forwarding is enabled.
1239
1240accept_source_route - INTEGER
1241        Accept source routing (routing extension header).
1242
1243        >= 0: Accept only routing header type 2.
1244        < 0: Do not accept routing header.
1245
1246        Default: 0
1247
1248autoconf - BOOLEAN
1249        Autoconfigure addresses using Prefix Information in Router
1250        Advertisements.
1251
1252        Functional default: enabled if accept_ra_pinfo is enabled.
1253                            disabled if accept_ra_pinfo is disabled.
1254
1255dad_transmits - INTEGER
1256        The amount of Duplicate Address Detection probes to send.
1257        Default: 1
1258
1259forwarding - INTEGER
1260        Configure interface-specific Host/Router behaviour.
1261
1262        Note: It is recommended to have the same setting on all
1263        interfaces; mixed router/host scenarios are rather uncommon.
1264
1265        Possible values are:
1266                0 Forwarding disabled
1267                1 Forwarding enabled
1268
1269        FALSE (0):
1270
1271        By default, Host behaviour is assumed.  This means:
1272
1273        1. IsRouter flag is not set in Neighbour Advertisements.
1274        2. If accept_ra is TRUE (default), transmit Router
1275           Solicitations.
1276        3. If accept_ra is TRUE (default), accept Router
1277           Advertisements (and do autoconfiguration).
1278        4. If accept_redirects is TRUE (default), accept Redirects.
1279
1280        TRUE (1):
1281
1282        If local forwarding is enabled, Router behaviour is assumed.
1283        This means exactly the reverse from the above:
1284
1285        1. IsRouter flag is set in Neighbour Advertisements.
1286        2. Router Solicitations are not sent unless accept_ra is 2.
1287        3. Router Advertisements are ignored unless accept_ra is 2.
1288        4. Redirects are ignored.
1289
1290        Default: 0 (disabled) if global forwarding is disabled (default),
1291                 otherwise 1 (enabled).
1292
1293hop_limit - INTEGER
1294        Default Hop Limit to set.
1295        Default: 64
1296
1297mtu - INTEGER
1298        Default Maximum Transfer Unit
1299        Default: 1280 (IPv6 required minimum)
1300
1301router_probe_interval - INTEGER
1302        Minimum interval (in seconds) between Router Probing described
1303        in RFC4191.
1304
1305        Default: 60
1306
1307router_solicitation_delay - INTEGER
1308        Number of seconds to wait after interface is brought up
1309        before sending Router Solicitations.
1310        Default: 1
1311
1312router_solicitation_interval - INTEGER
1313        Number of seconds to wait between Router Solicitations.
1314        Default: 4
1315
1316router_solicitations - INTEGER
1317        Number of Router Solicitations to send until assuming no
1318        routers are present.
1319        Default: 3
1320
1321use_tempaddr - INTEGER
1322        Preference for Privacy Extensions (RFC3041).
1323          <= 0 : disable Privacy Extensions
1324          == 1 : enable Privacy Extensions, but prefer public
1325                 addresses over temporary addresses.
1326          >  1 : enable Privacy Extensions and prefer temporary
1327                 addresses over public addresses.
1328        Default:  0 (for most devices)
1329                 -1 (for point-to-point devices and loopback devices)
1330
1331temp_valid_lft - INTEGER
1332        valid lifetime (in seconds) for temporary addresses.
1333        Default: 604800 (7 days)
1334
1335temp_prefered_lft - INTEGER
1336        Preferred lifetime (in seconds) for temporary addresses.
1337        Default: 86400 (1 day)
1338
1339max_desync_factor - INTEGER
1340        Maximum value for DESYNC_FACTOR, which is a random value
1341        that ensures that clients don't synchronize with each
1342        other and generate new addresses at exactly the same time.
1343        value is in seconds.
1344        Default: 600
1345
1346regen_max_retry - INTEGER
1347        Number of attempts before give up attempting to generate
1348        valid temporary addresses.
1349        Default: 5
1350
1351max_addresses - INTEGER
1352        Maximum number of autoconfigured addresses per interface.  Setting
1353        to zero disables the limitation.  It is not recommended to set this
1354        value too large (or to zero) because it would be an easy way to
1355        crash the kernel by allowing too many addresses to be created.
1356        Default: 16
1357
1358disable_ipv6 - BOOLEAN
1359        Disable IPv6 operation.  If accept_dad is set to 2, this value
1360        will be dynamically set to TRUE if DAD fails for the link-local
1361        address.
1362        Default: FALSE (enable IPv6 operation)
1363
1364        When this value is changed from 1 to 0 (IPv6 is being enabled),
1365        it will dynamically create a link-local address on the given
1366        interface and start Duplicate Address Detection, if necessary.
1367
1368        When this value is changed from 0 to 1 (IPv6 is being disabled),
1369        it will dynamically delete all address on the given interface.
1370
1371accept_dad - INTEGER
1372        Whether to accept DAD (Duplicate Address Detection).
1373        0: Disable DAD
1374        1: Enable DAD (default)
1375        2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
1376           link-local address has been found.
1377
1378force_tllao - BOOLEAN
1379        Enable sending the target link-layer address option even when
1380        responding to a unicast neighbor solicitation.
1381        Default: FALSE
1382
1383        Quoting from RFC 2461, section 4.4, Target link-layer address:
1384
1385        "The option MUST be included for multicast solicitations in order to
1386        avoid infinite Neighbor Solicitation "recursion" when the peer node
1387        does not have a cache entry to return a Neighbor Advertisements
1388        message.  When responding to unicast solicitations, the option can be
1389        omitted since the sender of the solicitation has the correct link-
1390        layer address; otherwise it would not have be able to send the unicast
1391        solicitation in the first place. However, including the link-layer
1392        address in this case adds little overhead and eliminates a potential
1393        race condition where the sender deletes the cached link-layer address
1394        prior to receiving a response to a previous solicitation."
1395
1396ndisc_notify - BOOLEAN
1397        Define mode for notification of address and device changes.
1398        0 - (default): do nothing
1399        1 - Generate unsolicited neighbour advertisements when device is brought
1400            up or hardware address changes.
1401
1402mldv1_unsolicited_report_interval - INTEGER
1403        The interval in milliseconds in which the next unsolicited
1404        MLDv1 report retransmit will take place.
1405        Default: 10000 (10 seconds)
1406
1407mldv2_unsolicited_report_interval - INTEGER
1408        The interval in milliseconds in which the next unsolicited
1409        MLDv2 report retransmit will take place.
1410        Default: 1000 (1 second)
1411
1412force_mld_version - INTEGER
1413        0 - (default) No enforcement of a MLD version, MLDv1 fallback allowed
1414        1 - Enforce to use MLD version 1
1415        2 - Enforce to use MLD version 2
1416
1417suppress_frag_ndisc - INTEGER
1418        Control RFC 6980 (Security Implications of IPv6 Fragmentation
1419        with IPv6 Neighbor Discovery) behavior:
1420        1 - (default) discard fragmented neighbor discovery packets
1421        0 - allow fragmented neighbor discovery packets
1422
1423icmp/*:
1424ratelimit - INTEGER
1425        Limit the maximal rates for sending ICMPv6 packets.
1426        0 to disable any limiting,
1427        otherwise the minimal space between responses in milliseconds.
1428        Default: 1000
1429
1430
1431IPv6 Update by:
1432Pekka Savola <pekkas@netcore.fi>
1433YOSHIFUJI Hideaki / USAGI Project <yoshfuji@linux-ipv6.org>
1434
1435
1436/proc/sys/net/bridge/* Variables:
1437
1438bridge-nf-call-arptables - BOOLEAN
1439        1 : pass bridged ARP traffic to arptables' FORWARD chain.
1440        0 : disable this.
1441        Default: 1
1442
1443bridge-nf-call-iptables - BOOLEAN
1444        1 : pass bridged IPv4 traffic to iptables' chains.
1445        0 : disable this.
1446        Default: 1
1447
1448bridge-nf-call-ip6tables - BOOLEAN
1449        1 : pass bridged IPv6 traffic to ip6tables' chains.
1450        0 : disable this.
1451        Default: 1
1452
1453bridge-nf-filter-vlan-tagged - BOOLEAN
1454        1 : pass bridged vlan-tagged ARP/IP/IPv6 traffic to {arp,ip,ip6}tables.
1455        0 : disable this.
1456        Default: 0
1457
1458bridge-nf-filter-pppoe-tagged - BOOLEAN
1459        1 : pass bridged pppoe-tagged IP/IPv6 traffic to {ip,ip6}tables.
1460        0 : disable this.
1461        Default: 0
1462
1463bridge-nf-pass-vlan-input-dev - BOOLEAN
1464        1: if bridge-nf-filter-vlan-tagged is enabled, try to find a vlan
1465        interface on the bridge and set the netfilter input device to the vlan.
1466        This allows use of e.g. "iptables -i br0.1" and makes the REDIRECT
1467        target work with vlan-on-top-of-bridge interfaces.  When no matching
1468        vlan interface is found, or this switch is off, the input device is
1469        set to the bridge interface.
1470        0: disable bridge netfilter vlan interface lookup.
1471        Default: 0
1472
1473proc/sys/net/sctp/* Variables:
1474
1475addip_enable - BOOLEAN
1476        Enable or disable extension of  Dynamic Address Reconfiguration
1477        (ADD-IP) functionality specified in RFC5061.  This extension provides
1478        the ability to dynamically add and remove new addresses for the SCTP
1479        associations.
1480
1481        1: Enable extension.
1482
1483        0: Disable extension.
1484
1485        Default: 0
1486
1487addip_noauth_enable - BOOLEAN
1488        Dynamic Address Reconfiguration (ADD-IP) requires the use of
1489        authentication to protect the operations of adding or removing new
1490        addresses.  This requirement is mandated so that unauthorized hosts
1491        would not be able to hijack associations.  However, older
1492        implementations may not have implemented this requirement while
1493        allowing the ADD-IP extension.  For reasons of interoperability,
1494        we provide this variable to control the enforcement of the
1495        authentication requirement.
1496
1497        1: Allow ADD-IP extension to be used without authentication.  This
1498           should only be set in a closed environment for interoperability
1499           with older implementations.
1500
1501        0: Enforce the authentication requirement
1502
1503        Default: 0
1504
1505auth_enable - BOOLEAN
1506        Enable or disable Authenticated Chunks extension.  This extension
1507        provides the ability to send and receive authenticated chunks and is
1508        required for secure operation of Dynamic Address Reconfiguration
1509        (ADD-IP) extension.
1510
1511        1: Enable this extension.
1512        0: Disable this extension.
1513
1514        Default: 0
1515
1516prsctp_enable - BOOLEAN
1517        Enable or disable the Partial Reliability extension (RFC3758) which
1518        is used to notify peers that a given DATA should no longer be expected.
1519
1520        1: Enable extension
1521        0: Disable
1522
1523        Default: 1
1524
1525max_burst - INTEGER
1526        The limit of the number of new packets that can be initially sent.  It
1527        controls how bursty the generated traffic can be.
1528
1529        Default: 4
1530
1531association_max_retrans - INTEGER
1532        Set the maximum number for retransmissions that an association can
1533        attempt deciding that the remote end is unreachable.  If this value
1534        is exceeded, the association is terminated.
1535
1536        Default: 10
1537
1538max_init_retransmits - INTEGER
1539        The maximum number of retransmissions of INIT and COOKIE-ECHO chunks
1540        that an association will attempt before declaring the destination
1541        unreachable and terminating.
1542
1543        Default: 8
1544
1545path_max_retrans - INTEGER
1546        The maximum number of retransmissions that will be attempted on a given
1547        path.  Once this threshold is exceeded, the path is considered
1548        unreachable, and new traffic will use a different path when the
1549        association is multihomed.
1550
1551        Default: 5
1552
1553pf_retrans - INTEGER
1554        The number of retransmissions that will be attempted on a given path
1555        before traffic is redirected to an alternate transport (should one
1556        exist).  Note this is distinct from path_max_retrans, as a path that
1557        passes the pf_retrans threshold can still be used.  Its only
1558        deprioritized when a transmission path is selected by the stack.  This
1559        setting is primarily used to enable fast failover mechanisms without
1560        having to reduce path_max_retrans to a very low value.  See:
1561        http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
1562        for details.  Note also that a value of pf_retrans > path_max_retrans
1563        disables this feature
1564
1565        Default: 0
1566
1567rto_initial - INTEGER
1568        The initial round trip timeout value in milliseconds that will be used
1569        in calculating round trip times.  This is the initial time interval
1570        for retransmissions.
1571
1572        Default: 3000
1573
1574rto_max - INTEGER
1575        The maximum value (in milliseconds) of the round trip timeout.  This
1576        is the largest time interval that can elapse between retransmissions.
1577
1578        Default: 60000
1579
1580rto_min - INTEGER
1581        The minimum value (in milliseconds) of the round trip timeout.  This
1582        is the smallest time interval the can elapse between retransmissions.
1583
1584        Default: 1000
1585
1586hb_interval - INTEGER
1587        The interval (in milliseconds) between HEARTBEAT chunks.  These chunks
1588        are sent at the specified interval on idle paths to probe the state of
1589        a given path between 2 associations.
1590
1591        Default: 30000
1592
1593sack_timeout - INTEGER
1594        The amount of time (in milliseconds) that the implementation will wait
1595        to send a SACK.
1596
1597        Default: 200
1598
1599valid_cookie_life - INTEGER
1600        The default lifetime of the SCTP cookie (in milliseconds).  The cookie
1601        is used during association establishment.
1602
1603        Default: 60000
1604
1605cookie_preserve_enable - BOOLEAN
1606        Enable or disable the ability to extend the lifetime of the SCTP cookie
1607        that is used during the establishment phase of SCTP association
1608
1609        1: Enable cookie lifetime extension.
1610        0: Disable
1611
1612        Default: 1
1613
1614cookie_hmac_alg - STRING
1615        Select the hmac algorithm used when generating the cookie value sent by
1616        a listening sctp socket to a connecting client in the INIT-ACK chunk.
1617        Valid values are:
1618        * md5
1619        * sha1
1620        * none
1621        Ability to assign md5 or sha1 as the selected alg is predicated on the
1622        configuration of those algorithms at build time (CONFIG_CRYPTO_MD5 and
1623        CONFIG_CRYPTO_SHA1).
1624
1625        Default: Dependent on configuration.  MD5 if available, else SHA1 if
1626        available, else none.
1627
1628rcvbuf_policy - INTEGER
1629        Determines if the receive buffer is attributed to the socket or to
1630        association.   SCTP supports the capability to create multiple
1631        associations on a single socket.  When using this capability, it is
1632        possible that a single stalled association that's buffering a lot
1633        of data may block other associations from delivering their data by
1634        consuming all of the receive buffer space.  To work around this,
1635        the rcvbuf_policy could be set to attribute the receiver buffer space
1636        to each association instead of the socket.  This prevents the described
1637        blocking.
1638
1639        1: rcvbuf space is per association
1640        0: rcvbuf space is per socket
1641
1642        Default: 0
1643
1644sndbuf_policy - INTEGER
1645        Similar to rcvbuf_policy above, this applies to send buffer space.
1646
1647        1: Send buffer is tracked per association
1648        0: Send buffer is tracked per socket.
1649
1650        Default: 0
1651
1652sctp_mem - vector of 3 INTEGERs: min, pressure, max
1653        Number of pages allowed for queueing by all SCTP sockets.
1654
1655        min: Below this number of pages SCTP is not bothered about its
1656        memory appetite. When amount of memory allocated by SCTP exceeds
1657        this number, SCTP starts to moderate memory usage.
1658
1659        pressure: This value was introduced to follow format of tcp_mem.
1660
1661        max: Number of pages allowed for queueing by all SCTP sockets.
1662
1663        Default is calculated at boot time from amount of available memory.
1664
1665sctp_rmem - vector of 3 INTEGERs: min, default, max
1666        Only the first value ("min") is used, "default" and "max" are
1667        ignored.
1668
1669        min: Minimal size of receive buffer used by SCTP socket.
1670        It is guaranteed to each SCTP socket (but not association) even
1671        under moderate memory pressure.
1672
1673        Default: 1 page
1674
1675sctp_wmem  - vector of 3 INTEGERs: min, default, max
1676        Currently this tunable has no effect.
1677
1678addr_scope_policy - INTEGER
1679        Control IPv4 address scoping - draft-stewart-tsvwg-sctp-ipv4-00
1680
1681        0   - Disable IPv4 address scoping
1682        1   - Enable IPv4 address scoping
1683        2   - Follow draft but allow IPv4 private addresses
1684        3   - Follow draft but allow IPv4 link local addresses
1685
1686        Default: 1
1687
1688
1689/proc/sys/net/core/*
1690        Please see: Documentation/sysctl/net.txt for descriptions of these entries.
1691
1692
1693/proc/sys/net/unix/*
1694max_dgram_qlen - INTEGER
1695        The maximum length of dgram socket receive queue
1696
1697        Default: 10
1698
1699
1700UNDOCUMENTED:
1701
1702/proc/sys/net/irda/*
1703        fast_poll_increase FIXME
1704        warn_noreply_time FIXME
1705        discovery_slots FIXME
1706        slot_timeout FIXME
1707        max_baud_rate FIXME
1708        discovery_timeout FIXME
1709        lap_keepalive_time FIXME
1710        max_noreply_time FIXME
1711        max_tx_data_size FIXME
1712        max_tx_window FIXME
1713        min_tx_turn_time FIXME
1714