- Socket handling revamped to use sockaddr_t.
- tinc can now tunnel over IPv6.
- Handle all addresses and subnets in network byte order.
Only convert them when they need to be printed.
- IPv6 subnets bigger than /128 now work.
- Use %s and strerror(errno) instead of %m.
- Transition to new node/vertex/connection structures
- Use new configuration handling everywhere
- Linux tun/tap device handling cleanup
- Start of IPv6 support in route.c
It compiles, but it won't link.
- Added seperate file for connection list handling
- Updating everything to use connlist, meta and subnet files
- Removed dependency on libgmp
- Lots of other stuff...
connection now has two hostnames: real_hostname (replacing the old),
and vpn_hostname. In those places where hostnames really aren't usefull
IP_ADDR_S has been replaced by %d.%d.%d.%d.
- Removed hostname lookup (it blocks, and you can always do it yourself)
- Reorganized debug levels (after hints from Axel M�ller):
0 Startup message and errors
1 Connection logging
2 Meta protocol information
3 Verbose meta protocol (includes copy of transmitted requests)
4 Packet information (logs transmission/errors of UDP packets)
5 Verbose packet information (every single byte, not implemented yet
to protect ourselves from filling up /var/log directories)
- Made log messages more consistent
A remote user could telnet to the tinc daemon and type only this line:
61 6 00000000/00000000:28f
This would deny any packets to be sent to other tinc networks (except
for to the hosts that run tincd's themselves). Solution is to skip
hosts in lookup_conn() that have not been activated yet.
Fixed potential conn_list table corruption:
If a new connection is accepted but a connection with the same subnet
would already exist in the connection list, the OLD connection is
terminated.
furthermore this version does NOT work yet because of a problem with
sending keys (these should be converted to base36 or something like that).
It is possible to telnet to the tinc daemon now and type some commands
by hand though :).
