Kea 2.7.1
libkea-asiodns - Kea I/O Library

The asiodns library is intended to provide an abstraction layer between Kea modules and asiolink library.

These DNS server and client routines are written using the "stackless coroutine" pattern invented by Chris Kohlhoff and described at: http://blog.think-async.com/2010/03/potted-guide-to-stackless-coroutines.html. This is intended to simplify development a bit, since it allows the routines to be written in a straightforward step-step-step fashion rather than as a complex chain of separate handler functions.

Coroutine objects (i.e., UDPServer, TCPServer and IOFetch) are objects with reentrant operator() members. When an instance of one of these classes is called as a function, it resumes at the position where it left off. Thus, an UDPServer can issue an asynchronous I/O call and specify itself as the handler object; when the call completes, the UDPServer carries on at the same position. As a result, the code can look as if it were using synchronous, not asynchronous I/O, providing some of the benefit of threading but with minimal switching overhead.

So, in simplified form, the behavior of a DNS Server is:

REENTER:
    while true:
        YIELD packet = read_packet
        FORK
        if not parent:
            break

    YIELD answer = DNSLookup(packet, this)
    response = DNSAnswer(answer)
    YIELD send(response)

At each "YIELD" point, the coroutine initiates an asynchronous operation, then pauses and turns over control to some other task on the ASIO service queue. When the operation completes, the coroutine resumes.

The DNSLookup and DNSAnswer define callback methods used by a DNS Server to communicate with the module that called it. They are abstract-only classes whose concrete implementations are supplied by the calling module.

The DNSLookup callback always runs asynchronously. Concrete implementations must be sure to call the server's "resume" method when it is finished.

In an authoritative server, the DNSLookup implementation would examine the query, look up the answer, then call "resume" (See the diagram in doc/auth_process.jpg).

In a recursive server, the DNSLookup implementation would initiate a DNSQuery, which in turn would be responsible for calling the server's "resume" method (See the diagram in doc/recursive_process.jpg).

A DNSQuery object is intended to handle resolution of a query over the network when the local authoritative data sources or cache are not sufficient. The plan is that it will make use of subsidiary DNSFetch calls to get data from\ particular authoritative servers, and when it has gotten a complete answer, it calls "resume".

In current form, however, DNSQuery is much simpler; it forwards queries to a single upstream resolver and passes the answers back to the client. It is constructed with the address of the forward server. Queries are initiated with the question to ask the forward server, a buffer into which to write the answer, and a pointer to the coroutine to be resumed when the answer has arrived. In simplified form, the DNSQuery routine is:

REENTER:
    render the question into a wire-format query packet
    YIELD send(query)
    YIELD response = read_packet
    server->resume

Currently, DNSQuery is only implemented for UDP queries. In future work it will be necessary to write code to fall back to TCP when circumstances require it.

Upstream Fetches

Upstream fetches (queries by the resolver on behalf of a client) are made using a slightly-modified version of the pattern described above.

Sockets

First, it will be useful to understand the class hierarchy used in the fetch logic:

     IOSocket
         |
    IOAsioSocket
         |
   +-----+-----+
   |           |

UDPSocket TCPSocket

IOSocket is a wrapper class for a socket and is used by the authoritative server code. It is an abstract base class, providing little more that the ability to hold the socket and to return the protocol in use.

Built on this is the IOAsioSocket, which adds the open, close, asyncSend and asyncReceive methods. This is a template class, which takes as template argument the class of the object that will be used as the callback when the asynchronous operation completes. This object can be of any type, but must include an operator() method with the signature:

operator()(boost::system::error_code ec, size_t length)

... the two arguments being the status of the completed I/O operation and the number of bytes transferred (in the case of the open method, the second argument will be zero).

Finally, the TCPSocket and UDPSocket classes provide the body of the asynchronous operations.

Fetch Sequence

The fetch is implemented by the IOFetch class which takes as argument the protocol to use. The sequence is:

REENTER:
    render the question into a wire-format query packet
    open()                           // Open socket and optionally connect
    if (! synchronous) {
        YIELD;
    }
    YIELD asyncSend(query)           // Send query
    do {
        YIELD asyncReceive(response) // Read response
    } while (! complete(response))
    close()                          // Drop connection and close socket
    server->resume

The open() method opens a socket for use. On TCP, it also makes a connection to the remote end. So under UDP the operation will complete immediately, but under TCP it could take a long time. One solution would befor the open operation to post an event to the I/O queue; then both cases could be regarded as being equivalent, with the completion being signalled by the posting of the completion event. However UDP is the most common case and that would involve extra overhead. So the open() returns a status indicating whether the operation completed asynchronously. If it did, the code yields back to the coroutine; if not the yield is bypassed.

The asynchronous send is straightforward, invoking the underlying ASIO function. Note that the address/port is supplied to both the open() and asyncSend() methods - it is used by the TCPSocket in open() and by the UDPSocket in asyncSend().

The asyncReceive() method issues an asynchronous read and waits for completion. The fetch object keeps track of the amount of data received so far and when the receive completes it calls a method on the socket to determine if the entire message has been received. This will always be the case for UDP. On TCP though, the message is preceded by a count field as several reads may be required to read all the data. The fetch loops until all the data is read.

Finally, the socket is closed and the server called to resume operation.