And reset to the async loop when sync request was finished, rescheduling the upstream->event.
Note that finished_event is scheduled against the async loop always.
* default_eventloop was prototyped in getdns_query and is still in there as my_eventloop
* It interfaces directly with the scheduling primitives of getdns.
* It can operate entirely from stack and does not have to do
any memory allocations or deallocations.
* Adapted configure.ac to allow libunbound to be linked with Windows
(with the removal of winsock_event.c we have no symbol clashed anymore)
* Added STUB_TCP_WOULDBLOCK return code in stub_resolving helper functions,
to anticipate dealing with edge triggered event loops (versus level triggered). (i.e. Windows)
We do this by associating a getdns_upstream object with the SSL object
handled by that upstream.
This allows us to collapse the verification callback code to a single
function.
Note that if we've agreed that fallback is ok, we are now willing to
accept *any* cert verification error, not just HOSTNAME_MISMATCH.
This is fine, because the alternative is falling back to cleartext,
which would be worse.
We also always set SSL_VERIFY_PEER, since we might as well try to do
so; we'll drop the verification error ourselves if we know we're OK
with falling back.
This is a parameter to the getdns_context that tells the context how
much to pad queries that go out over TLS.
It is not yet functional in this commit, but the idea is to pad each
outbound query over TLS to a multiple of the requested blocksize.
Because we only have a set amount of pre-allocated space for dynamic
options (MAXIMUM_UPSTREAM_OPTION_SPACE), we limit the maximum
padding blocksize.
This is a simplistic padding policy. Suggestions for improved padding
policies are welcome!
https://tools.ietf.org/html/draft-ietf-dnsop-edns-client-subnet-04
Using the above spec, an intermediate resolver may forward a chunk of
the client's IP address to the authoritative resolver.
Setting edns_client_subnet_private to a getdns_context in stub mode
will indicate to the next-hop recursive resolver that the client
wishes to keep their address information private.
- set and get functions are added.
- Existing transport functions retained for backwards compatibility.
- Basic combinations work as before, but underlying functional changes and cleanup are not complete yet...
- Context level options for timeouts and max_transactions_per_tcp_connection coming soon...
- timeouts are not being rescheduled on fallback
- several error cases are not being handled correctly (e.g. 8.8.8.8) and a user callback is not always called
- the fallback mechanism is not generic (specific to tls to tcp)
Willem Toorop
saradickinson
Daniel Kahn Gillmor