Provided by: npd6_1.1.0-2_amd64 bug

NAME

       npd6.conf - configuration file for neighbor proxy daemon ipv6 npd6

DESCRIPTION

       This  file  contains  parameters  which  affect  the manner of npd6's operation. The first
       section contain a few common parameters - you will need  to  change  these  to  suit  your
       situation. The second set are less likely to require changing and will typically work fine
       at their default value. In fact: if you change the second set and get it wrong,  npd6  may
       work  badly,  intermittently or not at all. Leave them alone unless you have a real reason
       to change them!

       All these parameters can be dauting. Fear not! Probably 95% of the folks who use npd6 have
       received  a  64-bit  prefix  from  their  ISP  and  want to have their box answer neighbor
       solicitations for the prefix on, maybe, interface eth0. And that's all they want. So  read
       a  bit  of  General Parameters below and set your one prefix and one interface and happily
       leave everything else alone... It'll work great for you.

   General parameters
       prefix = prefix[/mask]
              This  defines  the  prefix  which  will  be  matched  against   received   Neighbor
              Solicitations.  It  will often be a /64 supplied by an ISP, but this is by no means
              mandatory.

              Note that leading zeros are not required, but can be used. So
                     prefix=2a01:0123:0456:0007:
              is equivalent to
                     prefix=2a01:123:456:7:

              The prefix does require a terminating colon, e.g.
                     prefix=2a01:123:456:7:

              The /mask value is optional, and often not required. In many cases  where  npd6  is
              used,  the  user will have received a 64-bit prefix from their ISP. Simply add this
              as the prefix here and you'll be good to go!

              If the /mask value is not supplied, but the prefix given is not a 64-bit one,  then
              an  implicit  mask  is  used  of  the prefix's length rounded up to the next 16-bit
              boundary.  e.g.
                     prefix = 2000:11:22:33:8000::
              is implicitly euivalent to
                     prefix = 2000:11:22:33:8000::/80

              If a /mask value is used, then it  operates  in  the  traditional  (although  maybe
              unintuitive to some!) maner of an IP mask. For example:
                     prefix = 2000:11:22:33::/65
              would match 2000:11:22:33::0 thru 2000:11:22:33:7fff:ffff:ffff:ffff

              Whereas
                     prefix = 2000:11:22:33:8000::/65
              would match 2000:11:22:33:8000::0 thru 2000:11:22:33:ffff:ffff:ffff:ffff

              If  you  understand  masks,  then  they  are  there  and  available.  If you do not
              understand masks, just ignore this.

              Finally: a useful shorthand to match any address within  a  NS  on  the  designated
              interface is
                     prefix=0::/0
              However  its  use would be unusual and could lead to some unforseen consequences if
              not thought through. Use with care.

       interface = [interfacename]
              This defines the interface to which we will bind and listen. Note also that this is
              the  interface  which will, by default (see below), have the multicast flag enabled
              if required too.

              It will typically be one of the ethernet interfaces, e.g.
                     interface=eth0

       Note that there is no static  limit  to  how  many  interface/prefix  pairs  that  can  be
       specified. Also not fixed is how multiple pairs are configured: you can choose to list the
       interfaces first, followed by a list of prefixes. Or alternatively to pair them up in  the
       config  file  instead.  Note  however  that  there  must be an equal number of prefixes as
       interfaces. Every interface must have a prefix defined for it (if only a wildcard)

   Address blacklisting & whitelisting
       listtype = [none|black|white]
              This defines the  type  of  "listing"  (if  any)  to  be  performed:  whitelisting,
              blacklisting or none at all. What is "listing"? In general, a whitelist ONLY allows
              items pre-configured on the list to  be  permitted.  A  blacklist  allows  anything
              EXCEPT those items pre-configured on the list.

              If  black or white is specified then you should also then define at least 1 (and as
              many more as you require) addrlist and/or exprlist items.

              Note also that this config item must be before the addrlist or  exprlist  items  in
              the config file.

       addrlist= <ipv6 address>

              e.g.  addrlist = 2a01:0123:456:7::22 (All the usual formats are supported.)

              Defines the list of addresses to be either white- or black-listed.

       exprlist = <expression to match>

              e.g.  exprlist = (HOST);SEAN=(HOST&0xffff);((SEAN>0x99)&&(SEAN<0x124))

              This  example  will  (if black listing) refuse any NS that is for PREFIX::100 up to
              PREFIX::123. It any other PREFIX:: value will match. Or if whitelisting, will  only
              allow value in those ranges.

       listlogging = [off|on]
              If  on addresses which match a white/blacklist will be logged (even if debug is not
              enabled). If off matches will not be logged (unless debug is in use, when they will
              be logged regardless)

              Default: off

       List-type precedence

       Note that exprlist values take precedence over addrlist values (if both are specified)

       Behaviour with no black- or whitelisting

       This  is the default mode of operation and likely what most people will want. In this mode
       (the original mode before this feature was introduced) any Neighbor Solicitation  received
       with a target address which matches the prefix will be answered by an appropriate Neighbor
       Advertisement, irrespective of whether or  not  that  target  actually  exists  and/or  is
       reachable from here.

       Behaviour with blacklisting

       This mode of operation modifies how npd6 behaves: when a Neighbor Solicitation is received
       with a prefix which matches that configured, before a Neighbor Advertisement is  generated
       in  reply  the target is additionally checked against the list of addresses defined. If it
       matches one of those addresses then the Neighbor Solicitation is silently ignored  and  no
       response generated.

       Behaviour with whitelisting

       This mode of operation modifies how npd6 behaves: when a Neighbor Solicitation is received
       with a prefix which matches that configured, before a Neighbor Advertisement is  generated
       in reply the target is additionally checked against the list of addresses defined. Only if
       it matches one of those addresses is a Neighbor Solicitation is generated in response.

   Special parameters
       collectTargets = amount
              Optionally, npd6 can record a list of targets received  in  the  incoming  Neighbor
              Solicitations,  to  which it has replied. These can be displayed in the defined log
              by sending a USR2 signal npd6. e.g. kill -USR2 <pid of npd6>

              This parameter defines a limit to the number of addresses recorded. If the value is
              zero,  then  the  feature is disabled. This is the default behaviour. The parameter
              should be set to a suitably large but sane value. The only significant overhead  is
              memory  (for  storing  the  addresses)  Given  the  address  space  of IPv6 and the
              possibility of receiving very many different target  addresses  to  not  provide  a
              ceiling to this feature would potentially open npd6 up to a denial-of-service based
              upon memory depletion.

              In  a  typical  situation,  where  a  network  might  expect  to  receive  Neighbor
              Solicitations  for several dozen devices, there's likely no harm in setting this to
              100 or 1000. Just don't set it at 10,000,000 without thinking first. Memory is only
              allocated  as-required  - if you do set a high limit, npd6 does not preallocate the
              memory, but only use it as and if required.

       linkOption =  false|true
              Normal RFC-compliant behaviour requires that if npd6 is  replying  to  a  multicast
              Neighbor  Solicitation  it  must include the Target Link-Layer option in any reply.
              However if replying to a unicast Neighbor Solicitation this option is not required.
              Hence it is normally not used unless enabled here by changing the this to true

              Default: false

       ignoreLocal =  false|true
              If npd6 receives a Neighbor Solicitation for the global address of the interface to
              which it is bound, if set to true npd6 will ignore and not reply -  this  is  since
              ordinarily the kernel will itself automatically respond to such a situation.

              If  changed  from the default and set to false then such Neighbor Solcitations will
              be responded to - possibly resulting in duplicated Neighbor Advertisements.

              Default: true

       routerNA =  false|true
              Normal RFC-compliant behaviour requires that outgoing Neighbor Advertisements  have
              the  ROUTER flag set in them. This will be done unless disabled by changing this to
              false.

              Default: true

       maxHops =  [0-255]
              Outgoing Neighbor Advertisements will normally have their maximum hop value set  to
              255.  if set incorrectly no anomaly in the operation of npd6 will be seen - however
              Neghbor Advertisements generated by npd6 may be silently ignored by  the  receiving
              device. Do not change without good reason!

              Default: 255

       pollErrorLimit =  [0-255]
              This  is  to  cater for situations when an interface goes bad for a short time, but
              comes back in the end. e.g. bouncing an interface, or during  boot-up  when  things
              might take a second or three longer than expected to be ready and available.

              Set to 0 to disable the mechanism completely.

              Given  the  nature of the underlying mechanisms, if an interface goes away and does
              not come back, this threshold strictly speaking  is  the  minimum  number  of  poll
              failures we will wait for before permanently failing. However practically speaking,
              this value represents the minimum period in  seconds  which  we  will  wait  before
              taking this action. The maximum period is potentially twice this value.

              Note  that  if  this  value is set higher than 30, the mechanism may not operate as
              expected! This is implementation dependent. Best left alone, to be honest... 20  is
              a sensible default value!

              Default: 20

SEE ALSO

       npd6(8)

BUGS

       No known bugs - Please report them to npd6Project@gmail.com

AUTHOR

       Sean Groarke (sgroarke@gmail.com)