check valid address before UTP_Create()

[reardon]

While debugging other uTP code, I noticed that there is no addressing sanitizing, particularly for IPv6, when attempting to setup a uTP connection.

Seems like code in tr_netOpenPeerUTPSocket() should:

    assert( tr_address_is_valid( addr ) );

    if( !tr_address_is_valid_for_peers( addr, port ) )
        return -EINVAL;

prior to UTP_Create(), just like the sanity check in tr_netOpenPeerSocket()

This will avoid a ton of misfires due to bogus (mostly link-local) addresses we get from LTEP. Also, sanity check gives us a place to filter out Teredo and 6in4 peer connections (ie 2001:0::/32 and 2002::/16) should we want (see #4196)

[cfpp2p]

take a look at:

peer-io.c line 682

    assert( session );
    assert( tr_address_is_valid( addr ) );
    assert( torrentHash );

    if( utp )
        utp_socket = tr_netOpenPeerUTPSocket( session, addr, port, isSeed );

    if( !utp_socket ) {
        fd = tr_netOpenPeerSocket( session, addr, port, isSeed );
        dbgmsg( NULL, "tr_netOpenPeerSocket returned fd %d", fd );
    }

does that help sort any of this out for you?

[reardon]

Huh? Are you talking about the assert()?? Those are not relevant to release code, but regardless those do not check for the condition I'm referring to.

What follows the assert() is the (somewhat odd, different than uT) logic which tries uTP before trying regular TCP connections. Won't insulate us from bad addresses received via LTEP/PEX.

[jordan]

I like this idea. Added in r13628, copied from tr_netOpenPeerSocket()

[reardon]

I think it needs to return NULL instead of error, since the caller is not designed to handle error values. Alternatively, code elsewhere needs changes so that both netOpenPeer... routines share parallel logic, which they don't today.

[jordan]

D'oh! That was a terrible commit, thanks.

[x190]

Fixed in: r13629.