changelog

aodv协议的实现 使用linux环境下的c++编写

0.1:
* Initial release.

0.2:
* Fixed byte order problems in some headers.
* Corrected sequence number handling.
* Wait on reboot (DELETE_PERIOD).
* Draft 10 compliance.
* Gratuitous flag in RREQ for ICMP ECHO requests.
* Show node's own sequence number in routing table log.
* Made expanding ring search optional.
* ICMP Destination Host Unreachable to application on route lookup failure.
* Fixed pointer error in seek_list.c which caused seg-faults.
* Experimental Internet gateway support (more like a bad hack ;-).
* Fixed bugs when there was a hashing conflict in the routing table.
* Experimental support for receiving N hellos before treating a node as 
  a neighbor.

0.3:
* New timer code to fix obscure crashes (very stable now).
* Multiple network interfaces support.
* Fixed bug which caused some RERRs to not be sent.
* HELLO message jittering.
* Code is AODV-interop tested and verified.
* Updated README.
* New command line parsing, and options.
* Added command line option to disable wait on reboot (maybe Manel
  will stop whining about the 15 secs delay now ;-).

0.4:
* Fixed inconsistency in "--no-worb" option (was --disable-worb in --help).
* Fixes/changes for unidir and n-hello modes.
* Removed dependency on modified ip_queue kernel module. Now packets are 
  re-routed on the NF_IP_POST_ROUTING hook instead.
* Removed use of skb->protocol to detect IP packets in netfilter
  code. Apparently skb->protocol is not always set. This caused much
  grief, since some packets were not being processed by AODV.
* Fixed bug in RERR-code which made routes be invalidated twice.
* Fixed bug when TTL_START was set to last_hcnt=255 when doing RREQs.
* Squashed bug in packet_input.c which made routes not timeout when hello 
  messages stops being received.

0.5:
* Replaced "#ifdef DEBUG" with a DEBUG macro to make the code cleaner.
* Ported code to the NS-2 simulator. See README for more information.
* Fixed bug in hello_process_non_hello() which caused the hop count of 
  routes to not be updated.
* Changes to how the timer/event queue is aged to better fit NS.
* Fixed lots of RERR related bugs.
* Fixed so that expired routing table entries are never expunged before 
  PATH_TRAVERSAL_TIME when waiting for a RREP.
* Moved packet buffering to separate source file (packet_queue.c).

0.6: 
* Made it so that a check for queued packets to send is always done
  whenever a new route is added or updated.
* Implemented simple rate limiting by discarding outgoing control packets 
  in case the RATELIMIT is reached.
* Added new aodv_neighbor.c module to separate generic neighbor processing 
  from HELLO message processing. Now adding a neighbor is explicit and
  separate from for example HELLO processing.
* Code changed to conform with the new draft v11 terminology. Routing table 
  debug output also changed to reflect this.
* Fixed missing NULL pointer check for route pointer (line 320 of 
  packet_input.c) that potentially could cause a segfault.
* When using AODV-UU in NS, periodic routing table logging was not
  performed until the routing table was updated for the first time. Fixed.
* Compilation problems in NS under SuSE Linux ("undefined reference")
  have been fixed.
* NS support code updated to work with ns-2 version 2.1b9a.
* NS support code should now compile under Sun Solaris and other platforms
  which lack <endian.h>.
* Implemented local repair.
* Removed unnecessary flags in routing table.
* Fixed RREP forwarding so that precursor lists are updated correctly.
* Lots of code cleanups and restructuring.
* Implemented garbage collection of old packets in packet_queue.c.

0.7:
* Added new RREP waiting parameters to conform with AODV draft v13.
* Implemented a better RATELIMIT algorithm.
* Fixed broken make targets (install, all).
* Fixed bug that caused NO DELETE RERRs to not be sent. Thanks to McWood 
  <[e-mail]>.
* NS support code updated to work with ns-2 version 2.26.
* aodv_socket.c now use RAW UDP sockets only.
* New shared linked list implementation.
* Lots of cleanup... Please report bugs!
* Made link layer feedback a runtime option (in ns-2).

0.7.1: 
* Removed extra 8 bit "flags" in RREQ struct that was not supposed to
  be there. Thanks Krishna Prabhala <[e-mail]>, for
  discovering that.

0.7.2:
* Fixed small compilation bug when compiling against ns-2.

0.8:
* Fixed overflow bug in timeval_add_msec. Thanks to Stefan Book 
  <[e-mail]>.
* IP addresses are now stored as "struct in_addr".
* Fixed segmentation fault in timer queue when executing expired timers.
* Changed unknown sequence number behavior. 0 now means unknown. This
  is different from rfc3561, but functionally equivalent.
* Simulation supports hierarchical addressing.
* Included experimental gateway connectivity using tunnels (both ns-2
  and real world).
* Address locality can be implemented in locality.c
* Made sequence numbers rollover safe.
* HELLO messages and link layer feedback is now a runtime setting that
  can be toggled in the simulation script. AODV-UU also sets the
  correct time parameters automatically.
* Updated README.ns regarding the link layer feedback run-time option.
* Updated README.ns; it incorrectly stated that "make" and
  "make distclean" commands should be run with a leading "./".
* AODV-UU log files are renamed to aodvd.log and aodvd.rtlog.
* In ns-2, AODV-UU log files are now named by node number (not pseudo IP).
* AODV-UU no longer uses the mobilenode class in ns-2 that the ns-2
  bundled AODV uses. Instead, it implements a new node type called
  AODVNode which is similar to DSR's SRNode. This fixes a bug where
  routing table entries were not refreshed at the destination node since 
  packets were previously passed directly to the sink. See the README.ns 
  file.
* Fixed bug where HELLO timers were not updated when receiving data packets.
* Support for ns-2.27.
* In ns-2 (2.27), the RREP originator address is now logged in the trace 
  file.
* If CONFIG_GATEWAY is defined during compilation, AODV-UU will buffer
  full data packets in user space and use a RAW socket to send them. This is 
  to avoid some problems experienced with the netfilter queueing mechanism.