news

大名鼎鼎的路由器源码。程序分ZEBRA、OSPFRIP等3个包。程序框架采用一个路由协议一个进程的方式

GNU Zebra is not yet released, so this NEWS is about beta version.

* Changes in zebra-0.94

* Changes in zebra

** Do not listen other process's netlink message.

* Changes in bgpd

** "bgp log-neighbor-changes" is added.

** "set ip next-hop peer-address" is added.

** Community delete bug is fixed.

** Fix bug of router-id display

** Option parameter length bug is fixed.

* Chages in ospfd

** Point-to-Multipoint support.

** OSPF MD5 authentication bug is fixed.

** OSPF NSSA bug is fixed.

** NSM event schedule bug is fixed.

** Update Opaque LSA patch.

** When write queue becomes empty stop write timer.

* Changes in ospf6d

** Update to the latest Ohara's code.

** DR election bug is fixed.

** Update link-local address on interface creation

* Changes in vtysh

** Update for IPv6 handling

** Make all protocol DEFUN/ALIAS consistent

* Changes in lib

** Fix route-map problem

** Fix vty bug cause daemon crash.

* Changes in zebra-0.93

* Changes in bgpd

** Configuration is changed to new format.

* Changes in ospfd

** Crush bugs which reported on Zebra ML is fixed.

** Opaque LSA and TE LSA support is added by KDD R&D Laboratories,
   Inc.

* Chages in ospf6d

** Many bugs are fixed.

* Changes in zebra-0.92a

* Changes in bgpd

** Fix "^$" community list bug.

** Below command's Address Family specific configurations are added

  nexthop-self
  route-reflector-client
  route-server-client
  soft-reconfiguration inbound

* Changes in zebra

** Treat kernel type routes as EGP routes.

* Changes in zebra-0.92

** Overall security is improved.  Default umask is 0077.

* Changes in ripd

** If output interface is in simple password authentication mode,
substruct one from rtemax.

* Changes in bgpd

** IPv4 multicast and IPv6 unicast configuration is changed to so
called new config.  All of AFI and SAFI specific configuration is
moved to "address-family" node.  When you have many IPv6 only
configuration, you will see many "no neighbor X:X::X:X activate" line
in your configuration to disable IPv4 unicast NLRI exchange.  In that
case please use "no bgp default ipv4-unicast" command to suppress the
output.  Until zebra-0.93, old config is still left for compatibility.

Old config
==========
router bgp 7675
 bgp router-id 10.0.0.1
 redistribute connected
 network 192.168.0.0/24
 neighbor 10.0.0.2 remote-as 7675
 ipv6 bgp network 3ffe:506::/33
 ipv6 bgp network 3ffe:1800:e800::/40
 ipv6 bgp aggregate-address 3ffe:506::/32
 ipv6 bgp redistribute connected
 ipv6 bgp neighbor 3ffe:506:1000::2 remote-as 1

New config
==========
router bgp 7675
 bgp router-id 10.0.0.1
 network 192.168.0.0/24
 redistribute connected
 neighbor 10.0.0.2 remote-as 7675
 neighbor 3ffe:506:1000::2 remote-as 1
 no neighbor 3ffe:506:1000::2 activate
!
 address-family ipv6
  network 3ffe:506::/33
  network 3ffe:1800:e800::/40
  aggregate-address 3ffe:506::/32
  redistribute connected
  neighbor 3ffe:506:1000::2 activate
 exit-address-family

* Changes in ospfd

** Internal interface treatment is changed.  Now ospfd can handle
multiple IP address for an interface.

** Redistribution of loopback interface's address works fine.

* Changes in zebra-0.91

** --enable-oldrib configure option is removed.

** HAVE_IF_PSEUDO part is removed.  Same feature is now supported by
default.

* Changes in ripd

** When redistributed route is withdrawn, perform poisoned reverse.

* Changes in zebra

** When interface's address is removed, kernel route pointing out to
the address is removed.

** IPv6 RIB is now based upon new RIB code.

** zebra can handle same connected route to one interface.

** New command for interface address.  Currently this commands are
only supported on GNU/Linux with netlink interface.

"ip address A.B.C.D secondary"
"ip address A.B.C.D label LABEL"

* Changes in bgpd

** BGP flap dampening bugs are fixed.

** BGP non-blocking TCP connection bug is fixed.

** "show ip bgp summary" shows AS path and community entry number.

** New commands have been added.
  "show ip bgp cidr-only"
  "show ip bgp ipv4 (unicast|multicast) cidr-only"
  "show ip bgp A.B.C.D/M longer-prefixes"
  "show ip bgp ipv4 (unicast|multicast) A.B.C.D/M longer-prefixes"
  "show ipv6 bgp X:X::X:X/M longer-prefixes"
  "show ipv6 mbgp X:X::X:X/M longer-prefixes"

** IPv6 IBGP nexthop change is monitored.

** Unknown transitive attribute is passed with partial flag bit on.

* Changes in ospfd

** Fix bug of LSA MaxAge flood.

** Fix bug of NSSA codes.

* Changes in zebra-0.90

** From this beta release, --enable-unixdomain and --enable-newrib
becomes default.  So both options are removed from configure.in.  To
revert old behavior please specify below option.

--enable-tcp-zebra # TCP/IP socket is used for protocol daemon and zebra.
--enable-oldrib    # Turn on old RIB implementation.

Old RIB implementation will be removed in zebra-0.91.

** From this beta release --enable-multipath is supported.  This
option is only effective on GNU/Linux kernel with
CONFIG_IP_ADVANCED_ROUTER and CONFIG_IP_ROUTE_MULTIPATH is set.

--enable-multipath=ARG  # ARG must be digit.  When ARG is 0 unlimit multipath number.

** From this release we do not include guile files.

* Changes in lib

** newlist.[ch] is merged with linklist.[ch].

** Now Zebra works on MacOS X public beta.

** Access-list can have remark.  "access-list WORD remark LINE" define
remark for specified access-list.

** Key of key-chain is sorted by it's idetifier value.

** prefix-list rule is slightly changed.  The rule of "len <= ge-value
<= le-value" is changed to "len < ge-value <= le-value".

** According to above prefix-list rule change, add automatic
conversion function of an old rule. ex.) 10.0.0.0/8 ge 8 -> 10.0.0.0/8
le 32

** SMUX can handle SNMP trap.

** In our event library, event thread is executed before any other
thread like timer, read and write event.

** Robust method for writing configuration file and recover from
backing up config file.

** Display "end" at the end of configuration.

** Fix memory leak in vtysh_read().

** Fix memroy leak about access-list and prefix-list name.

* Changes in zebra

** UNIX domain socket server of zebra protocol is added.

** Fix PointoPoint interface network bug.  The destination network
should be installed into routing table instead of local network.

** Metric value is reflected to kernel routing table.

** "show ip route" display uptime of RIP,OSPF,BGP routes.

** New RIB implementation is added.

Now we have enhanced RIB (routing information base) implementation in
zebra.  New RIB has many new features and fixed some bugs which exist
in old RIB code.

*** Static route with distance value

    Static route can be specified with administrative distance.  The
  distance value 255 means it is not installed into the kernel.
  Default value of distance for static route is 1.
  
    ip route A.B.C.D/M A.B.C.D <1-255>
    ip route A.B.C.D/M IFNAME <1-255>
  
    If the least distance value's route's nexthop are unreachable,
  select the least distance value route which has reachable nexthop is
  selected.
  
    ip route 0.0.0.0/0 10.0.0.1
    ip route 0.0.0.0/0 11.0.0.1 2
  
    In this case, when 10.0.0.1 is unreachable and 11.0.0.1 is
  reachable.  The route with nexthop 11.0.0.1 will be installed into
  forwarding table.
  
    zebra> show ip route
    S>* 0.0.0.0/0 [2/0] via 11.0.0.1
    S   0.0.0.0/0 [1/0] via 10.0.0.1 inactive
  
    If the nexthop is unreachable "inactive" is displayed.  You can
  specify any string to IFNAME.  There is no need of the interface is
  there when you configure the route.
  
    ip route 1.1.1.1/32 ppp0
  
  When ppp0 comes up, the route is installed properly.

*** Multiple nexthop routes for one prefix

    Multiple nexthop routes can be specified for one prefix.  Even the
  kernel support only one nexthop for one prefix user can configure
  multiple nexthop.
  
    When you configure routes like below, prefix 10.0.0.1 has three
  nexthop.
  
    ip route 10.0.0.1/32 10.0.0.2
    ip route 10.0.0.1/32 10.0.0.3
    ip route 10.0.0.1/32 eth0
  
    If there is no route to 10.0.0.2 and 10.0.0.3.  And interface eth0
    is reachable, then the last route is installed into the kernel.
  
    zebra> show ip route
    S>  10.0.0.1/32 [1/0] via 10.0.0.2 inactive
                          via 10.0.0.3 inactive
      *                   is directly connected, eth0
  
    '*' means this nexthop is installed into the kernel. 

*** Multipath (more than one nexthop for one prefix) can be installed into the kernel.

    When the kernel support multipath, zebra can install multipath
  routes into the kernel.  Before doing that please make it sure that
  setting --enable-multipath=ARG to configure script.  ARG must be digit
  value.  When specify 0 to ARG, there is no limitation of the number
  of the multipath.  Currently only GNU/Linux with netlink interface is
  supported.
  
    ip route 10.0.0.1/32 10.0.0.2
    ip route 10.0.0.1/32 10.0.0.3
    ip route 10.0.0.1/32 eth0
  
    zebra> show ip route
    S>* 10.0.0.1/32 [1/0] via 10.0.0.2
      *                   via 10.0.0.3
                          is directly connected, eth0

*** Kernel message delete installed route.

    After zebra install static or dynamic route into the kernel.
  
    R>* 0.0.0.0/0 [120/3] via 10.0.0.1
  
    If you delete this route outside zebra, old zebra does not reinstall
  route again.  Now the route is re-processed and properly reinstall the
  static or dynamic route into the kernel.

** GNU/Linux netlink socket handling is improved to fix race condition
between kernel message and user command responce.

* Changes in bgpd

** Add show neighbor's routes command.

  "show ip bgp neighbors (A.B.C.D|X:X::X:X) routes"
  "show ip bgp ipv4 (unicast|multicast) neighbors (A.B.C.D|X:X::X:X) routes"
  "show ipv6 bgp neighbors (A.B.C.D|X:X::X:X) routes"
  "show ipv6 mbgp neighbors (A.B.C.D|X:X::X:X) routes"

** BGP passive peer support problem is fixed.

** Redistributed IGP nexthop is passed to BGP nexthop.

** On multiaccess media, if the nexthop is reachable nexthop is passed
as it is.

** Remove zebra-0.88 compatibility commands.

  "match ip prefix-list WORD"
  "match ipv6 prefix-list WORD"

  Instead of above please use below commands.

  "match ip address prefix-list WORD"
  "match ipv6 address prefix-list WORD"

** Fix bug of holdtimer is not reset when bgp cleared.

** "show ip bgp summary" display peer establish/drop count.

** Change "match ip next-hop" argument from IP address to access-list
name.

** When "bgp enforce-first-as" is enabled, check EBGP peer's update
has it's AS number in the first AS number in AS sequence.

** New route-map command "set community-delete COMMUNITY-LIST" is
added.  Community matched the CoMMUNITY-LIST is removed from the
community.

** BGP-MIB implementation is finished.

** When BGP connection comes from unconfigured IP address, close
socket immediately.

** Do not compare router ID when the routes comes from EBGP peer.
When originator ID is same, take shorter cluster-list route.  If
cluster-list is same take smaller IP address neighbor's route.

** Add "bgp bestpath as-path ignore" command.  When this option is
set, do not concider AS path length when route selection.

** Add "bgp bestpath compare-routerid".  When this option is set,
compare router ID when the routes comes from EBGP peer.

** Add "bgp deterministic-med" process.

** BGP flap dampening feature is added.

** When IBGP nexthop is changed, it is reflected to RIB.

** Change "neighbor route-refresh" command to "neighbor capability
route-refresh".

* Changes in ripd

** Change "match ip next-hop" argument from IP address to access-list
name.

** "no ip rip (send|receive)" command accept version number argument.

** Memory leak related classfull network generation is fixed.

** When a route is in garbage collection process (invalid with metric
16) and a router receives the same route with valid metric then route
was not installed into zebra rib, but only into ripd rib. Moreover ,
it will never get into zebra rib, because ripd wrongly assumes it's
already there.

* Change in ospfd

** Fix bug of refreshing default route.

** --enable-nssa turn on undergoing NSSA feature.

** Fix bug of Hello packet's option is not properly set when interface
comes up.

** Reduce unconditional logging.

** Add nexthop to OSPF path only when it is not there.

** When there is no DR on network (suppose you have only one router
with interface priority 0). It's router LSA does not contain the link
information about this network.

** When you change a priority of interface from/to 0
ISM_NeighborChange event should be scheduled in order to elect new
DR/BDR on the network.

** When we add some LSA into retransmit list we need to check whether
the present old LSA in retransmit list is not more recent than the new
one.

** In states Loading and Full the slave must resend its last Database
Description packet in response to duplicate Database Description
packets received from the master.  For this reason the slave must wait
RouterDeadInterval seconds before freeing the last Database
Description packet. Reception of a Database Description packet from
the master after this interval will generate a SeqNumberMismatch
neighbor event. RFC2328 Section 10.8

** Virtual link can not configured in stub area.

** Clear a ls_upd_queue queue of the interface when interface goes
down.

** "no router ospf" unregister redistribution requests from zebra.

** New command for virtual-link configuration is added.

  "area A.B.C.D virtual-link A.B.C.D"
  "area A.B.C.D virtual-link A.B.C.D hello-interval <1-65535> retransmit-interval <3-65535> transmit-delay <1-65535> dead-interval <1-65535>"
  "area A.B.C.D virtual-link A.B.C.D hello-interval <1-65535> retransmit-interval <3-65535> transmit-delay <1-65535> dead-interval <1-65535> authentication-key AUTH_KEY"
  "area A.B.C.D virtual-link A.B.C.D authentication-key AUTH_KEY"
  "area A.B.C.D virtual-link A.B.C.D hello-interval <1-65535> retransmit-interval <3-65535> transmit-delay <1-65535> dead-interval <1-65535> message-digest-key <1-255> md5 KEY"
  "area A.B.C.D virtual-link A.B.C.D message-digest-key <1-255> md5 KEY"