Description of the getdns API

Paul Hoffman, Editor

Document version: "getdns September 2014"

This document describes a modern asynchronous DNS API. This new API is intended to be useful to application developers and operating system distributors as a way of making all types of DNS information easily available in many types of programs. The major features of this new API are:

• Full support for event-driven programming
• Supports DNSSEC in multiple ways
• Mirroring of the resolution in getaddrinfo()
• Easily supports all RRtypes, even those yet to be defined

There is more background into the design and future goals of this API later in this document.

This document was discussed on the getdns-api mailing list; future versions of the API might be discussed there as well. (If you want to contact the editor off-list, please send mail to paul.hoffman@vpnc.org.)

1. The getdns Async Functions

The API has four async functions:

• getdns_address for doing getaddrinfo()-like address lookups
• getdns_hostname for doing getnameinfo()-like name lookups
• getdns_service for getting results from SRV lookups
• getdns_general for looking up any type of DNS record

1.1 getdns_general()

context

A pointer to the DNS context that is to be used with this call. DNS contexts are described later in this document. Note that a context must be created before calling the function.

*name

This is a null-terminted string consisting of an ASCII-based domain name to be looked up. The values here follow the rules in section 2.1 of RFC 4343 to allow non-ASCII octets and special characters in labels.

request_type

Specifies the RRtype for the query; the RRtype numbers are listed in the IANA registry. For example, to get the NS records, request_type would be 2. The API also has defined macros for most of the RRtypes by name; the definition names all start with "GETDNS_RRTYPE_". For example, to get the NS records, you can also set the request_type to GETDNS_RRTYPE_NS. (The full list of request types is always here.)

*extensions

Specifies the extensions for this request; the value may be NULL if there are no extensions. See the section below for information on how to specify the extensions used for a request.

*userarg

A void* that is passed to the function, which the funciton returns to the callback function untouched. userarg can be used by the callback function for any user-specific data needed. This can be NULL.

*transaction_id

A pointer to a value that is filled in by the function to identify the callback being made. This pointer can be NULL, in which case it is ignored and no value is assigned. The getdns_cancel_callback() function uses the transaction_id to determine which callback is to be cancelled. If the function fails, transaction_id is set to 0.

*callbackfn

A pointer to a callback function that is defined by the application. Typically, the callback function will do all the processing on the results from the API. The parameters of the callback are defined below. This really needs to be a pointer to a function (and not something like NULL); otherwise, the results are unpredictable.

The async getdns functions return GETDNS_RETURN_GOOD if the call was properly formatted. It returns GETDNS_RETURN_BAD_DOMAIN_NAME if the API determines that the name passed to the function was bad, GETDNS_RETURN_BAD_CONTEXT if the context has internal deficiencies, GETDNS_RETURN_NO_SUCH_EXTENSION if one or more extensions do not exist, or GETDNS_RETURN_EXTENSION_MISFORMAT if the contents of one or more of the extensions is incorrect. All of the return values are given later in this document.

1.2 getdns_address()

There are three critical differences between getdns_address() and getdns_general() beyond the missing request_type argument:

• In getdns_address(), the name argument can only take a host name.
• You do not need to include a return_both_v4_and_v6 extension with the call in getdns_address(): it will always return both IPv4 and IPv6 addresses.
• getdns_address() always uses all of namespaces from the context (to better emulate getaddrinfo()), while getdns_general() only uses the DNS namespace.

1.3 getdns_hostname()

The address is given as a getdns_dict data structure (defined below). The list must have two names: address_type (whose value is a bindata; it is currently either "IPv4" or "IPv6" (which are case-sensitive)) and address_data (whose value is a bindata).

1.4 getdns_service()

name must be a domain name for an SRV lookup; the call returns the relevant SRV information for the name.

1.5 Callback Functions for getdns

A call to the async getdns functions typically returns before any network or file I/O occurs. After the API marshalls all the needed information, it calls the callback function that was passed by the application. The callback function might be called at any time, even before the calling function has returned. The API guarantees that the callback will be called exactly once unless the calling function returned an error, in which case the callback function is never called.

The getdns calling function calls the callback with the parameters defined as follows:

context

The DNS context that was used in the calling function. See below for a description of the basic use of contexts, and later for more advanced use.

callback_type

Supplies the reason for the callback. See below for the codes and reasons.

*response

A response object with the response data. This is described below. The application is responsible for cleaning up the response object with getdns_dict_destroy.

*userarg

Identical to the *userarg passed to the calling function.

transaction_id

The transaction identifier that was assigned by the calling function.

The following are the values for callback_type.

GETDNS_CALLBACK_COMPLETE

The response has the requested data in it

GETDNS_CALLBACK_CANCEL

The calling program cancelled the callback; response is NULL

GETDNS_CALLBACK_TIMEOUT

The requested action timed out; response is filled in with empty structures

GETDNS_CALLBACK_ERROR

The requested action had an error; response is NULL

1.6 Setting Up The DNS Context

Calls to getdns functions require a DNS context, which is a group of API settings that affect how DNS calls are made. For most applications, a default context is sufficient.

To create a new DNS context, use the function:

The call to getdns_context_create immediately returns a context that can be used with other API calls; that context contains the API's default values. Most applications will want set_from_os set to 1.

To clean up the context, including cleaning up all outstanding transactions that were called using this context, use the function:

When getdns_context_destroy() returns, the application knows that all outstanding transactions associated with this context will have been called; callbacks that had not been called before getdns_context_destroy() was called will be called with a callback_type of GETDNS_CALLBACK_CANCEL. getdns_context_destroy() returns after all of the needed cleanup is done and callbacks are made.

1.7 Canceling a Callback

To cancel an outstanding callback, use the following function.

This causes the API to call the callback with a callback_type of GETDNS_CALLBACK_CANCEL if the callback for this transaction_id has not already been called. This will cancel the callback regardless of what the original call was doing (such as in the middle of a DNS request, while DNSSEC validation is happening, and so on). The callback code for cancellation should clean up any memory related to the identified call, such as to deallocate the memory for the userarg. getdns_cancel_callback() may return immediately, even before the callback finishes its work and returns. Calling getdns_cancel_callback() with a transaction_id of a callback that has already been called or an unknown transaction_id returns GETDNS_RETURN_UNKNOWN_TRANSACTION; otherwise, getdns_cancel_callback() returns GETDNS_RETURN_GOOD.

1.8 Event-driven Programs

Event-driven programs (sometimes called "async programs") require an event base and event loop (among other things). Different event libraries have different structures or the event base. Because of this, there is no standard method to set the event base in the DNS API: those are all added as extensions. The API is distributed as a core package and one or more sets of extensions to align with event libraries. It is mandatory to use one of the extension functions to set the event base in the DNS context; this is required before calling any event-driven calls like the getdns functions.

Each implementation of the DNS API will specify an extension function that tells the DNS context which event base is being used. For example, one implementation of this API that uses the libevent event library might name this function "getdns_extension_set_libevent_base()" while another might name it "getdns_extension_set_eventbase_for_libevent()"; the two extension functions could have very different calling patterns and return values. Thus, the application developer must read the API documentation (not just this design document) in order to determine what extension function to use to tell the API the event base to use.

The structure of a typical event-driven application might look like the following pseudocode. The code in italics is specific to the event mechanism.

Includes for one or more regular C libraries
An include for the getdns library specific to the event library you use
Definition of your callback function
    Get the DNS data from the allocated pointer
    Process that data
    Check for errors
Definition of main()
    Create context
    Set up your event base
    Point the context to your event base
    Set up the getdns call arguments
    Make the getdns call
    Check if the getdns return is good
    Destroy the context
    Exit

The API does not have direct support for a polling interface. Instead, the callback interface is specifically designed to allow an application that wants to process results in polling instead of in callbacks to be able to create its own polling interface fairly trivially. Such a program would create a data structure for the calls, including their transaction_id and userarg. The userarg could be the polling data structure or have a pointer to it. The application would have just one callback function for all requests, and that function would copy the response into application memory, update the data structure based on the transaction_id index, and return from the callback. The polling code could then check the data structure for any updates at its leisure.

1.9 Calling the API Synchronously (Without Events)

There are functions parallel to the four getdns async functions, except that there is no callback. That is, when an application calls one of these synchronous functions, the API gathers all the required information and then returns the result. The value returned is exactly the same as the response returned in the callback if you had used the async version of the function.

When you are done with the data in the response, use the following function so that the API can free the memory from its internal pool.

2. Data structures in the API

The API returns data structures. The data structure is not a representational language like JSON: it is really just a data structure. Data structures can have four types of members:

• list is an ordered list, like JSON and Python lists. The members of the list can be any of the four data types.
• dict is a name-value pair, like a JSON object or Python dict. The name is a string literal, and the value can be any of the four data types. The order of the name-value pairs in a dict is not important.
• int is an integer compatible with uint32_t.
• bindata is a struct to hold binary data. It is defined as { size_t size; uint8_t *binary_stuff; }.

The API comes with helper functions to get data from the list and dict data types:

All of these helper getter functions return GETDNS_RETURN_GOOD if the call is successful. The list functions will return GETDNS_RETURN_NO_SUCH_LIST_ITEM if the index argument is out of range; the dict functions will return GETDNS_RETURN_NO_SUCH_DICT_NAME if the name argument doesn't exist in the dict. The functions also return GETDNS_RETURN_WRONG_TYPE_REQUESTED if the requested data type doesn't match the contents of the indexed argument or name.

This document uses graphical representations of data structures. It is important to note that this is only a graphical representation; the brackets, commas, quotation marks, comments, and so on are not part of the data. Also, this document uses macro names instead of some of the int arguments; of course, the data structures have the actual int in them.

The helper getter functions return references to getdns_dict, getdns_list and getdns_bindata data structures. The user must not directly destroy these retrieved "child" data structures; instead, they will automatically be destroyed when the containing "parent" data structure is destroyed. Because of this, retrieved "child" data structures cannot be used any more after the containing "parent" data structure has been destroyed.

2.1 Creating Data Structures

Some of the features of the API require that you create your own data structures to be used in arguments passed to the API. For example, if you want to use any extensions for the calling functions, you need to create a dict. The requisite functions are:

Lists are extended with the getdns_list_set_ calls with the index set to the size of the list (such as 0 for an empty list). Dicts are extended with the getdns_dict_set_ calls with the name set to a name that does not yet exist. Name-value pairs are removed with getdns_dict_remove_name().

These helper setter functions return GETDNS_RETURN_GOOD if the call is successful. The functions return GETDNS_RETURN_WRONG_TYPE_REQUESTED if the requested data type doesn't match the contents of the indexed argument or name. The list functions will return GETDNS_RETURN_NO_SUCH_LIST_ITEM if the index argument is higher than the length of the list. getdns_dict_remove_name() will return GETDNS_RETURN_NO_SUCH_DICT_NAME if the name argument doesn't exist in the dict.

The helper setter functions store copies of the given "child" values. It is the responsibility of the caller to dispose of the original values.

3. Extensions

Extensions are dict data structures. The names in the dict are the names of the extensions. The definition of each extension describes the value associated with the name. For most extensions, it is an on-off boolean, and the value is GETDNS_EXTENSION_TRUE. (There is not currently a good reason to specify an extension name and give it a value of GETDNS_EXTENSION_FALSE, but that is allowed by the API.)

For example, to create a dict for extensions and specify the extension to only return results that have been validated with DNSSEC, you might use:

/* . . . */
getdns_dict * this_extensions = getdns_dict_create();
this_ret = getdns_dict_set_int(this_extensions, "dnssec_return_only_secure", GETDNS_EXTENSION_TRUE);
/* . . . Do some processing with the extensions and results . . . */
/* Remember to clean up memory*/
getdns_dict_destroy(this_extensions);

The extensions described in this section are are:

• dnssec_return_status
• dnssec_return_only_secure
• dnssec_return_validation_chain
• return_both_v4_and_v6
• add_opt_parameters
• add_warning_for_bad_dns
• specify_class
• return_call_debugging

3.1 Extensions for DNSSEC

If an application wants the API to do DNSSEC validation for a request, it must set one or more DNSSEC-related extensions. Note that the default is for none of these extensions to be set and the API will not perform DNSSEC. Note that getting DNSSEC results can take longer in a few circumstances.

To return the DNSSEC status for each DNS record in the replies_tree list, use the dnssec_return_status extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to cause the returned status to have the name dnssec_status (an int) added to the other names in the record's dict ("header", "question", and so on). The values for that name are GETDNS_DNSSEC_SECURE, GETDNS_DNSSEC_BOGUS, GETDNS_DNSSEC_INDETERMINATE, and GETDNS_DNSSEC_INSECURE. Thus, a reply might look like:

    {     # This is the first reply
      "dnssec_status": GETDNS_DNSSEC_INDETERMINATE,
      "header": { "id": 23456, "qr": 1, "opcode": 0, ... },
      . . .

If instead of returning the status, you want to only see secure results, use the dnssec_return_only_secure extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to cause only records that the API can validate as secure with DNSSEC to be returned in the replies_tree and replies_full lists. No additional names are added to the dict of the record; the change is that some records might not appear in the results. When this context option is set, if the API receives DNS replies but none are determined to be secure, the error code at the top level of the response object is GETDNS_RESPSTATUS_NO_SECURE_ANSWERS.

Applications that want to do their own validation will want to have the DNSSEC-related records for a particular response. Use the dnssec_return_validation_chain extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to cause a set of additional DNSSEC-related records needed for validation to be returned in the response object. This set comes as validation_chain (a list) at the top level of the response object. This list includes all resource record dicts for all the resource records (DS, DNSKEY and their RRSIGs) that are needed to perform the validation from the root up. Thus, a reply might look like:

{     # This is the response object
  "validation_chain":
  [ { "name": <bindata for .>,
      "type": GETDNS_RRTYPE_DNSKEY,
      "rdata": { "flags": 256, . . . },
      . . . 
    },
    { "name": <bindata for .>,
      "type": GETDNS_RRTYPE_DNSKEY,
      "rdata": { "flags": 257, . . . },
      . . .
    },
    { "name": <bindata for .>,
      "type": GETDNS_RRTYPE_RRSIG,
      "rdata": { "signers_name": <bindata for .>,
                 "type_covered": GETDNS_RRTYPE_DNSKEY,
                 . . .
               },
    },
    { "name": <bindata for com.>,
      "type": GETDNS_RRTYPE_DS,
      . . .
    },
    { "name": <bindata for com.>,
      "type": GETDNS_RRTYPE_RRSIG
      "rdata": { "signers_name": <bindata for .>,
                 "type_covered": GETDNS_RRTYPE_DS,
                 . . .
               },
      . . .
    },
    { "name": <bindata for com.>,
      "type": GETDNS_RRTYPE_DNSKEY
      "rdata": { "flags": 256, . . . },
      . . .
    },
    { "name": <bindata for com.>,
      "type": GETDNS_RRTYPE_DNSKEY
      "rdata": { "flags": 257, . . . },
      . . .
    },
    { "name": <bindata for com.>,
      "type": GETDNS_RRTYPE_RRSIG
      "rdata": { "signers_name": <bindata for com.>,
                 "type_covered": GETDNS_RRTYPE_DNSKEY,
                 . . .
               },
      . . .
    },
    { "name": <bindata for example.com.>,
      "type": GETDNS_RRTYPE_DS,
      . . .
    },
    { "name": <bindata for example.com.>,
      "type": GETDNS_RRTYPE_RRSIG
      "rdata": { "signers_name": <bindata for com.>,
                 "type_covered": GETDNS_RRTYPE_DS,
                 . . .
               },
      . . .
    },
    { "name": <bindata for example.com.>,
      "type": GETDNS_RRTYPE_DNSKEY
      "rdata": { "flags": 257, ... },
      . . .
    },
    . . .
  ]
  "replies_tree":
  [
  . . .

If a request is using a context in which stub resolution is set, and that request also has any of the dnssec_return_status, dnssec_return_only_secure, or dnssec_return_validation_chain extensions specified, the API will not perform the request and will instead return an error of GETDNS_RETURN_DNSSEC_WITH_STUB_DISALLOWED.

3.2 Returning Both IPv4 and IPv6 Responses

Many applications want to get both IPv4 and IPv6 addresses in a single call so that the results can be processed together. The getdns_address and getdns_address_sync functions are able to do this automatically. If you are using the getdns_general or getdns_general_sync function, you can enable this with the return_both_v4_and_v6 extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to cause the results to be the lookup of either A or AAAA records to include any A and AAAA records for the queried name (otherwise, the extension does nothing). These results are expected to be used with Happy Eyeballs systems that will find the best socket for an application.

3.3 Setting Up OPT Resource Records

For lookups that need an OPT resource record in the Additional Data section, use the add_opt_parameters extension. The extension's value (a dict) contains the parameters; these are described in more detail in RFC 6891. They are:

• maximum_udp_payload_size (an int), a value between 512 and 65535; if not specified, this defaults to those from the DNS context
• extended_rcode (an int), a value between 0 and 255; if not specified, this defaults to those from the DNS context
• version (an int), a value between 0 and 255; if not specified, this defaults to 0
• do_bit (an int), a value between 0 and 1; if not specified, this defaults to those from the DNS context
• options (a list) contains dicts for each option to be specified. Each list time contains two names: option_code (an int) and option_data (a bindata). The API marshalls the entire set of options into a properly-formatted RDATA for the resource record.

It is very important to note that the OPT resource record specified in the add_opt_parameters extension might not be the same the one that the API sends in the query. For example, if the application also includes any of the DNSSEC extensions, the API will make sure that the OPT resource record sets the resource record appropriately, making the needed changes to the settings from the add_opt_parameters extension.

The use of this extension can conflict with the values in the DNS context. For example, the default for an OS might be a maximum payload size of 65535, but the extension might specify 1550. In such a case, the API will honor the values stated in the extension, but will honor the values from the DNS context if values are not given in the extension.

3.4 Getting Warnings for Responses that Violate the DNS Standard

To receive a warning if a particular response violates some parts of the DNS standard, use the add_warning_for_bad_dns extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to cause each reply in the replies_tree to contain an additional name, bad_dns (a list). The list is zero or more ints that indicate types of bad DNS found in that reply. The list of values is:

GETDNS_BAD_DNS_CNAME_IN_TARGET

A DNS query type that does not allow a target to be a CNAME pointed to a CNAME

GETDNS_BAD_DNS_ALL_NUMERIC_LABEL

One or more labels in a returned domain name is all-numeric; this is not legal for a hostname

GETDNS_BAD_DNS_CNAME_RETURNED_FOR_OTHER_TYPE

A DNS query for a type other than CNAME returned a CNAME response

3.5 Using Other Class Types

The vast majority of DNS requests are made with the Internet (IN) class. To make a request in a different DNS class, use, the specify_class extension. The extension's value (an int) contains the class number. Few applications will ever use this extension.

3.6 Extensions Relating to the API

An application might want to see debugging information for queries such as the length of time it takes for each query to return to the API. Use the return_call_debugging extension. The extension's value (an int) is set to GETDNS_EXTENSION_TRUE to add the name call_debugging (a list) to the top level of the response object. Each member of the list is a dict that represents one call made for the call to the API. Each member has the following names:

• query_name (a bindata) is the name that was sent
• query_type (an int) is the type that was queried for
• query_to (a bindata) is the address to which the query was sent
• start_time (a bindata) is the time the query started in milliseconds since the epoch, represented as a uint64_t
• end_time (a bindata) is the time the query was received in milliseconds since the epoch, represented as a uint64_t
• entire_reply (a bindata) is the entire response received
• dnssec_result (an int) is the DNSSEC status, or GETDNS_DNSSEC_NOT_PERFORMED if DNSSEC validation was not performed

4. Response Data from Queries

The callback function contains a pointer to a response object. A response object is always a dict. The response object always contains at least three names: replies_full (a list) and replies_tree (a list), and status (an int). replies_full is a list of DNS replies (each is bindata) as they appear on the wire. replies_tree is a list of DNS replies (each is a dict) with the various part of the reply parsed out. status is a status code for the query.

Because the API might be extended in the future, a response object might also contain names other than replies_full, replies_tree, and status. Similarly, any of the dicts described here might be extended in later versions of the API. Thus, an application using the API must not assume that it knows all possible names in a dict.

The following lists the status codes for response objects. Note that, if the status is that there are no responses for the query, the lists in replies_full and replies_tree will have zero length.

GETDNS_RESPSTATUS_GOOD

At least one response was returned

GETDNS_RESPSTATUS_NO_NAME

Queries for the name yielded all negative responses

GETDNS_RESPSTATUS_ALL_TIMEOUT

All queries for the name timed out

GETDNS_RESPSTATUS_NO_SECURE_ANSWERS

The context setting for getting only secure responses was specified, and at least one DNS response was received, but no DNS response was determined to be secure through DNSSEC.

GETDNS_RESPSTATUS_ALL_BOGUS_ANSWERS

The context setting for getting only secure responses was specified, and at least one DNS response was received, but all received responses for the requested name were bogus.

The top level of replies_tree can optionally have the following names: canonical_name (a bindata), intermediate_aliases (a list), answer_ipv4_address (a bindata), answer_ipv6_address (a bindata), and answer_type (an int).

• The value of canonical_name is the name that the API used for its lookup. It is in FQDN presentation format.
• The values in the intermediate_aliases list are domain names from any CNAME or unsynthesized DNAME found when resolving the original query. The list might have zero entries if there were no CNAMEs in the path. These may be useful, for example, for name comparisons when following the rules in RFC 6125.
• The value of answer_ipv4_address and answer_ipv6_address are the addresses of the server from which the answer was received.
• The value of answer_type is the type of name service that generated the response. The values are:

GETDNS_NAMETYPE_DNS

Normal DNS (RFC 1035)

GETDNS_NAMETYPE_WINS

The WINS name service (some reference needed)

If the call was getdns_address or getdns_address_sync, the top level of replies_tree has an additional name, just_address_answers (a list). The value of just_address_answers is a list that contains all of the A and AAAA records from the answer sections of any of the replies, in the order they appear in the replies. Each item in the list is a dict with at least two names: address_type (whose value is a bindata; it is currently either "IPv4" or "IPv6") and address_data (whose value is a bindata). Note that the dnssec_return_only_secure extension affects what will appear in the just_address_answers list. Also note if later versions of the DNS return other address types, those types will appear in this list as well.

The API can make service discovery through SRV records easier. If the call was getdns_service or getdns_service_sync, the top level of replies_tree has an additional name, srv_addresses (a list). The list is ordered by priority and weight based on the weighting algorithm in RFC 2782, lowest priority value first. Each element of the list is dict has at least two names: port and domain_name. If the API was able to determine the address of the target domain name (such as from its cache or from the Additional section of responses), the dict for an element will also contain address_type (whose value is a bindata; it is currently either "IPv4" or "IPv6") and address_data (whose value is a bindata). Note that the dnssec_return_only_secure extension affects what will appear in the srv_addresses list.

4.1 Structure of DNS replies_tree

The names in each entry in the the replies_tree list for DNS responses include header (a dict), question (a dict), answer (a list), authority (a list), and additional (a list), corresponding to the sections in the DNS message format. The answer, authority, and additional lists each contain zero or more dicts, with each dict in each list representing a resource record.

The names in the header dict are all the fields from Section 4.1.1. of RFC 1035. They are: id, qr, opcode, aa, tc, rd, ra, z, rcode, qdcount, ancount, nscount, and arcount. All are ints.

The names in the question dict are the three fields from Section 4.1.2. of RFC 1035: qname (a bindata), qtype (an int), and qclass (an int).

Resource records are a bit different than headers and question sections in that the RDATA portion often has its own structure. The other names in the resource record dicts are name (a bindata), type (an int), class (an int), ttl (an int) and rdata (a dict); there is no name equivalent to the RDLENGTH field. The OPT resource record does not have the class and the ttl name, but in stead provides udp_payload_size (an int), extended_rcode (an int), version (an int), do (an int), and z (an int).

The rdata dict has different names for each response type. There is a complete list of the types defined in the API. For names that end in "-obsolete" or "-unknown", the bindata is the entire RDATA field. For example, the rdata for an A record has a name ipv4_address (a bindata); the rdata for an SRV record has the names priority (an int), weight (an int), port (an int), and target (a bindata).

Each rdata dict also has a rdata_raw field (a bindata). This is useful for types not defined in this version of the API. It also might be of value if a later version of the API allows for additional parsers. Thus, doing a query for types not known by the API still will return a result: an rdata with just a rdata_raw.

It is expected that later extensions to the API will give some DNS types different names. It is also possible that later extensions will change the names for some of the DNS types listed above.

For example, a response to a getdns_address() call for www.example.com would look something like this:

{     # This is the response object
  "replies_full": [ <bindata of the first response>, <bindata of the second response> ],
  "just_address_answers":
  [
    {
      "address_type": <bindata of "IPv4">,
      "address_data": <bindata of 0x0a0b0c01>,
    },
    {
      "address_type": <bindata of "IPv6">,
      "address_data": <bindata of 0x33445566334455663344556633445566>
    }
  ],
  "canonical_name": <bindata of "www.example.com">,
  "answer_type": GETDNS_NAMETYPE_DNS,
  "intermediate_aliases": [],
  "replies_tree":
  [
    {     # This is the first reply
      "header": { "id": 23456, "qr": 1, "opcode": 0, ... },
      "question": { "qname": <bindata of "www.example.com">, "qtype": 1, "qclass": 1 },
      "answer":
      [
        {
          "name": <bindata of "www.example.com">,
          "type": 1,
          "class": 1,
          "ttl": 33000,
          "rdata":
          {
            "ipv4_address": <bindata of 0x0a0b0c01>
            "rdata_raw": <bindata of 0x0a0b0c01>
          }
        }
      ],
      "authority":
      [
        {
          "name": <bindata of "ns1.example.com">,
          "type": 1,
          "class": 1,
          "ttl": 600,
          "rdata":
          {
            "ipv4_address": <bindata of 0x65439876>
            "rdata_raw": <bindata of 0x65439876>
          }
        }
      ]
      "additional": [],
      "canonical_name": <bindata of "www.example.com">,
      "answer_type": GETDNS_NAMETYPE_DNS
    },
    {     # This is the second reply
      "header": { "id": 47809, "qr": 1, "opcode": 0, ... },
      "question": { "qname": <bindata of "www.example.com">, "qtype": 28, "qclass": 1 },
      "answer":
      [
        {
          "name": <bindata of "www.example.com">,
          "type": 28,
          "class": 1,
          "ttl": 1000,
          "rdata":
          {
            "ipv6_address": <bindata of 0x33445566334455663344556633445566>
            "rdata_raw": <bindata of 0x33445566334455663344556633445566>
          }
       }
      ],
      "authority": [  # Same as for other record... ]
      "additional": [],
    },
  ]
}

In DNS responses, domain names are treated special. RFC 1035 describes a form of name compression that requires that the entire record be available for analysis. The API deals with this by converting compressed names into full names when returning names in the replies_tree. This conversion happens for qname in question; name in the answer, authority, and additional; and in domain names in the data in names under rdata where the response type is AFSDB, CNAME, MX, NS, PTR, RP, RT, and SOA.

4.2 Converting Domain Names

Names in DNS fields are stored in a fashion very different from the normal presentation format normally used in applications. The DNS format is described in the first paragraph in Section 3.1 of RFC 1035; the presentation format here is a null-terminated string with interior dots. These helper functions only work with names in the DNS format that are not compressed. They are useful for converting domain names in the replies_tree to and from the FQDN presentation format.

getdns_convert_dns_name_to_fqdn() converts a domain name in DNS format to the presentation format. For example, the hex sequence 03 77 77 77 07 65 78 61 6d 70 6c 65 03 63 6f 6d 00 would be converted to "www.example.com". getdns_convert_fqdn_to_dns_name() does the reverse: it converts a null-terminated string in FQDN format to bytes in DNS format.

The returned values are allocated with the default system allocator, namely malloc. The caller is responsible of disposing these allocations with free.

5. Additional Definitions and Descriptions

5.1 A Few Needed Definitions

5.2 Return Codes

The return codes for all the functions are:

GETDNS_RETURN_GOOD

Good

GETDNS_RETURN_GENERIC_ERROR

Generic error

GETDNS_RETURN_BAD_DOMAIN_NAME

Badly-formed domain name in first argument

GETDNS_RETURN_BAD_CONTEXT

The context has internal deficiencies

GETDNS_RETURN_CONTEXT_UPDATE_FAIL

Did not update the context

GETDNS_RETURN_UNKNOWN_TRANSACTION

An attempt was made to cancel a callback with a transaction_id that is not recognized

GETDNS_RETURN_NO_SUCH_LIST_ITEM

A helper function for lists had an index argument that was too high.

GETDNS_RETURN_NO_SUCH_DICT_NAME

A helper function for dicts had a name argument that for a name that is not in the dict.

GETDNS_RETURN_WRONG_TYPE_REQUESTED

A helper function was supposed to return a certain type for an item, but the wrong type was given.

GETDNS_RETURN_NO_SUCH_EXTENSION

A name in the extensions dict is not a valid extension.

GETDNS_RETURN_EXTENSION_MISFORMAT

One or more of the extensions have a bad format.

GETDNS_RETURN_DNSSEC_WITH_STUB_DISALLOWED

A query was made with a context that is using stub resolution and a DNSSEC extension specified.

GETDNS_RETURN_MEMORY_ERROR

Unable to allocate the memory required.

GETDNS_RETURN_INVALID_PARAMETER

A required parameter had an invalid value.

5.3 Types of RDATA Returned in the API

The names in the rdata dicts in replies are:

A (1)

ipv4_address (a bindata)

NS (2)

nsdname (a bindata)

MD (3)

madname (a bindata)

MF (4)

madname (a bindata)

CNAME (5)

cname (a bindata)

SOA (6)

mname (a bindata), rname (a bindata), serial (an int), refresh (an int), refresh (an int), retry (an int), and expire (an int)

MB (7)

madname (a bindata)

MG (8)

mgmname (a bindata)

MR (9)

newname (a bindata)

NULL (10)

anything (a bindata)

WKS (11)

address (a bindata), protocol (an int), and bitmap (a bindata)

PTR (12)

ptrdname (a bindata)

HINFO (13)

cpu (a bindata) and os (a bindata)

MINFO (14)

rmailbx (a bindata) and emailbx (a bindata)

MX (15)

preference (an int) and exchange (a bindata)

TXT (16)

txt_strings (a list) which contains zero or more bindata elements that are text strings

RP (17)

mbox_dname (a bindata) and txt_dname (a bindata)

AFSDB (18)

subtype (an int) and hostname (a bindata)

X25 (19)

psdn_address (a bindata)

ISDN (20)

isdn_address (a bindata) and sa (a bindata)

RT (21)

preference (an int) and intermediate_host (a bindata)

NSAP (22)

nsap (a bindata)

SIG (24)

sig_obsolete (a bindata)

KEY (25)

key_obsolete (a bindata)

PX (26)

preference (an int), map822 (a bindata), and mapx400 (a bindata)

GPOS (27)

longitude (a bindata), latitude (a bindata), and altitude (a bindata)

AAAA (28)

ipv6_address (a bindata)

LOC (29)

loc_obsolete (a bindata)

NXT (30)

nxt_obsolete (a bindata)

EID (31)

eid_unknown (a bindata)

NIMLOC (32)

nimloc_unknown (a bindata)

SRV (33)

priority (an int), weight (an int), port (an int), and target (a bindata)

ATMA (34)

format (an int) and address (a bindata)

NAPTR (35)

order (an int), preference (an int), flags (a bindata), service (a bindata), regexp (a bindata), and replacement (a bindata).

KX (36)

preference (an int) and exchanger (a bindata)

CERT (37)

type (an int), key_tag (an int), algorithm (an int), and certificate_or_crl (a bindata)

A6 (38)

a6_obsolete (a bindata)

DNAME (39)

target (a bindata)

SINK (40)

sink_unknown (a bindata)

OPT (41)

options (a list). Each element of the options list is a dict with two names: option_code (an int) and option_data (a bindata).

APL (42)

apitems (a list). Each element of the apitems list is a dict with four names: address_family (an int), prefix (an int), n (an int), and afdpart (a bindata)

DS (43)

key_tag (an int), algorithm (an int), digest_type (an int), and digest (a bindata)

SSHFP (44)

algorithm (an int), fp_type (an int), and fingerprint (a bindata)

IPSECKEY (45)

algorithm (an int), gateway_type (an int), precedence (an int), gateway, and public_key (a bindata)

RRSIG (46)

type_covered (an int), algorithm (an int), labels (an int), original_ttl (an int), signature_expiration (an int), signature_inception (an int), key_tag (an int), signers_name (a bindata), and signature (a bindata)

NSEC (47)

next_domain_name (a bindata) and type_bit_maps (a bindata)

DNSKEY (48)

flags (an int), protocol (an int), algorithm (an int), and public_key (a bindata)

DHCID (49)

dhcid_opaque (a bindata)

NSEC3 (50)

hash_algorithm (an int), flags (an int), iterations (an int), salt (a bindata), next_hashed_owner_name (a bindata), and type_bit_maps (a bindata)

NSEC3PARAM (51)

hash_algorithm (an int), flags (an int), iterations (an int), and salt (a bindata)

TLSA (52)

certificate_usage (an int), selector (an int), matching_type (an int), and certificate_association_data (a bindata).

HIP (55)

pk_algorithm (an int), hit (a bindata), public_key (a bindata), and rendezvous_servers (a list) with each element a bindata with the dname of the rendezvous_server.

NINFO (56)

ninfo_unknown (a bindata)

RKEY (57)

rkey_unknown (a bindata)

TALINK (58)

talink_unknown (a bindata)

CDS (59)

cds_unknown (a bindata)

CDNSKEY (60)

cdnskey_unknown (a bindata)

OPENPGPKEY (61)

openpgpkey_unknown (a bindata)

SPF (99)

text (a bindata)

UINFO (100)

uinfo_unknown (a bindata)

UID (101)

uid_unknown (a bindata)

GID (102)

gid_unknown (a bindata)

UNSPEC (103)

unspec_unknown (a bindata)

NID (104)

preference (an int) and node_id (a bindata)

L32 (105)

preference (an int) and locator32 (a bindata)

L64 (106)

preference (an int) and locator64 (a bindata)

LP (107)

preference (an int) and fqdn (a bindata)

EUI48 (108)

eui48_address (a bindata)

EUI64 (109)

eui64_address (a bindata)

TKEY (249)

algorithm (a bindata), inception (an int), expiration (an int), mode (an int), error (an int), key_data (a bindata), and other_data (a bindata)

TSIG (250)

algorithm (a bindata), time_signed (a bindata), fudge (an int), mac (a bindata), original_id (an int), error (an int), and other_data (a bindata)

MAILB (253)

mailb-unknown (a bindata)

MAILA (254)

maila-unknown (a bindata)

URI (256)

priority (an int), weight (an int), and target (a bindata)

CAA (257)

flags (an int), tag (a bindata), and value (a bindata)

TA (32768)

ta_unknown (a bindata)

DLV (32769)

Identical to DS (43)

6. Examples

This section gives examples of code that calls the API to do many common tasks. The purpose of the code here is to give application developers a quick hands-on demo of using the API.

Note that the examples here all use getdns_libevent.h as the include that will call in the API code as well as calling in libevent as the event library. They also use getdns_context_set_libevent_base() as the name of the function to set the event base in the DNS context. If you are using a different event library, you will of course use a different #include at the beginning of your code, and a different name for the event base function.

6.1 Get Both IPv4 and IPv6 Addresses for a Domain Name Using Quick Results

This is an example of a common call to getdns_address().

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <getdns_libevent.h>

#define UNUSED_PARAM(x) ((void)(x))

/* Set up the callback function, which will also do the processing of the results */
void this_callbackfn(getdns_context *this_context,
                     getdns_callback_type_t this_callback_type,
                     getdns_dict *this_response, 
                     void *this_userarg,
                     getdns_transaction_t this_transaction_id)
{
    UNUSED_PARAM(this_userarg);  /* Not looking at the userarg for this example */
    UNUSED_PARAM(this_context);  /* Not looking at the context for this example */
    getdns_return_t this_ret;  /* Holder for all function returns */
    if (this_callback_type == GETDNS_CALLBACK_COMPLETE)  /* This is a callback with data */
    {
        /* Be sure the search returned something */
        uint32_t this_error;
        this_ret = getdns_dict_get_int(this_response, "status", &this_error);  // Ignore any error
        if (this_error != GETDNS_RESPSTATUS_GOOD)  // If the search didn't return "good"
        {
            fprintf(stderr, "The search had no results, and a return value of %d. Exiting.\n", this_error);
            getdns_dict_destroy(this_response);
            return;
        }
        getdns_list * just_the_addresses_ptr;
        this_ret = getdns_dict_get_list(this_response, "just_address_answers", &just_the_addresses_ptr);
        if (this_ret != GETDNS_RETURN_GOOD)  // This check is really not needed, but prevents a compiler error under "pedantic"
        {
            fprintf(stderr, "Trying to get the answers failed: %d\n", this_ret);
            getdns_dict_destroy(this_response);
            return;
        }
        size_t num_addresses;
        this_ret = getdns_list_get_length(just_the_addresses_ptr, &num_addresses);  // Ignore any error
        /* Go through each record */
        for ( size_t rec_count = 0; rec_count < num_addresses; ++rec_count )
        {
            getdns_dict * this_address;
            this_ret = getdns_list_get_dict(just_the_addresses_ptr, rec_count, &this_address);  // Ignore any error
            /* Just print the address */
            getdns_bindata * this_address_data;
            this_ret = getdns_dict_get_bindata(this_address, "address_data", &this_address_data); // Ignore any error
            char *this_address_str = getdns_display_ip_address(this_address_data);
            printf("The address is %s\n", this_address_str);
            free(this_address_str);
        }
    }
    else if (this_callback_type == GETDNS_CALLBACK_CANCEL)
        fprintf(stderr, "The callback with ID %"PRIu64" was cancelled. Exiting.\n", this_transaction_id);
    else
        fprintf(stderr, "The callback got a callback_type of %d. Exiting.\n", this_callback_type);
    getdns_dict_destroy(this_response);
}

int main()
{
    /* Create the DNS context for this call */
    getdns_context *this_context = NULL;
    getdns_return_t context_create_return = getdns_context_create(&this_context, 1);
    if (context_create_return != GETDNS_RETURN_GOOD)
    {
        fprintf(stderr, "Trying to create the context failed: %d", context_create_return);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    /* Create an event base and put it in the context using the unknown function name */
    struct event_base *this_event_base;
    this_event_base = event_base_new();
    if (this_event_base == NULL)
    {
        fprintf(stderr, "Trying to create the event base failed.\n");
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    (void)getdns_extension_set_libevent_base(this_context, this_event_base);
    /* Set up the getdns call */
    const char * this_name  = "www.example.com";
    char* this_userarg = "somestring"; // Could add things here to help identify this call
    getdns_transaction_t this_transaction_id = 0;

    /* Make the call */
    getdns_return_t dns_request_return = getdns_address(this_context, this_name,
        NULL, this_userarg, &this_transaction_id, this_callbackfn);
    if (dns_request_return == GETDNS_RETURN_BAD_DOMAIN_NAME)
    {
        fprintf(stderr, "A bad domain name was used: %s. Exiting.\n", this_name);
        event_base_free(this_event_base);
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    else
    {
        /* Call the event loop */
        int dispatch_return = event_base_dispatch(this_event_base);
        UNUSED_PARAM(dispatch_return);
        // TODO: check the return value above
    }
    /* Clean up */
    event_base_free(this_event_base);
    getdns_context_destroy(this_context);
    /* Assuming we get here, leave gracefully */
    exit(EXIT_SUCCESS);
}

6.2 Get IPv4 and IPv6 Addresses for a Domain Name

This example is similar to the previous one, except that it retrieves more information than just the addresses, so it traverses the replies_tree. In this case, it gets both the addresses and their TTLs.

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <getdns_libevent.h>

#define UNUSED_PARAM(x) ((void)(x))

/* Set up the callback function, which will also do the processing of the results */
void this_callbackfn(getdns_context *this_context,
                     getdns_callback_type_t this_callback_type,
                     getdns_dict *this_response, 
                     void *this_userarg,
                     getdns_transaction_t this_transaction_id)
{
    UNUSED_PARAM(this_userarg);  /* Not looking at the userarg for this example */
    UNUSED_PARAM(this_context);  /* Not looking at the context for this example */
    getdns_return_t this_ret;  /* Holder for all function returns */
    if (this_callback_type == GETDNS_CALLBACK_COMPLETE)  /* This is a callback with data */
    {
        /* Be sure the search returned something */
        uint32_t this_error;
        this_ret = getdns_dict_get_int(this_response, "status", &this_error);  // Ignore any error
        if (this_error != GETDNS_RESPSTATUS_GOOD)  // If the search didn't return "good"
        {
            fprintf(stderr, "The search had no results, and a return value of %d. Exiting.\n", this_error);
            getdns_dict_destroy(this_response);
            return;
        }
        /* Find all the answers returned */
        getdns_list * these_answers;
        this_ret = getdns_dict_get_list(this_response, "replies_tree", &these_answers);
        if (this_ret == GETDNS_RETURN_NO_SUCH_DICT_NAME)
        {
            fprintf(stderr, "Weird: the response had no error, but also no replies_tree. Exiting.\n");
            getdns_dict_destroy(this_response);
            return;
        }
        size_t num_answers;
        this_ret = getdns_list_get_length(these_answers, &num_answers);
        /* Go through each answer */
        for ( size_t rec_count = 0; rec_count < num_answers; ++rec_count )
        {
            getdns_dict * this_record;
            this_ret = getdns_list_get_dict(these_answers, rec_count, &this_record);  // Ignore any error
            /* Get the answer section */
            getdns_list * this_answer;
            this_ret = getdns_dict_get_list(this_record, "answer", &this_answer);  // Ignore any error
            /* Get each RR in the answer section */
            size_t num_rrs;
            this_ret = getdns_list_get_length(this_answer, &num_rrs);
            for ( size_t rr_count = 0; rr_count < num_rrs; ++rr_count )
            {
                getdns_dict *this_rr = NULL;
                this_ret = getdns_list_get_dict(this_answer, rr_count, &this_rr);  // Ignore any error
                /* Get the RDATA */
                getdns_dict * this_rdata = NULL;
                this_ret = getdns_dict_get_dict(this_rr, "rdata", &this_rdata);  // Ignore any error
                /* Get the RDATA type */
                uint32_t this_type;
                this_ret = getdns_dict_get_int(this_rr, "type", &this_type);  // Ignore any error
                /* If it is type A or AAAA, print the value */
                if (this_type == GETDNS_RRTYPE_A)
                {
                    getdns_bindata * this_a_record = NULL;
                    this_ret = getdns_dict_get_bindata(this_rdata, "ipv4_address", &this_a_record);
                    if (this_ret == GETDNS_RETURN_NO_SUCH_DICT_NAME)
                    {
                        fprintf(stderr, "Weird: the A record at %d in record at %d had no address. Exiting.\n",
                            (int) rr_count, (int) rec_count);
                        getdns_dict_destroy(this_response);
                        return;
                    }
                    char *this_address_str = getdns_display_ip_address(this_a_record);
                    printf("The IPv4 address is %s\n", this_address_str);
                    free(this_address_str);
                }
                else if (this_type == GETDNS_RRTYPE_AAAA)
                {
                    getdns_bindata * this_aaaa_record = NULL;
                    this_ret = getdns_dict_get_bindata(this_rdata, "ipv6_address", &this_aaaa_record);
                    if (this_ret == GETDNS_RETURN_NO_SUCH_DICT_NAME)
                    {
                        fprintf(stderr, "Weird: the AAAA record at %d in record at %d had no address. Exiting.\n",
                            (int) rr_count, (int) rec_count);
                        getdns_dict_destroy(this_response);
                        return;
                    }
                    char *this_address_str = getdns_display_ip_address(this_aaaa_record);
                    printf("The IPv6 address is %s\n", this_address_str);
                    free(this_address_str);
                }
            }
        }
    }
    else if (this_callback_type == GETDNS_CALLBACK_CANCEL)
        fprintf(stderr, "The callback with ID %"PRIu64" was cancelled. Exiting.\n", this_transaction_id);
    else
        fprintf(stderr, "The callback got a callback_type of %d. Exiting.\n", this_callback_type);
    getdns_dict_destroy(this_response);
}

int main()
{
    /* Create the DNS context for this call */
    getdns_context *this_context = NULL;
    getdns_return_t context_create_return = getdns_context_create(&this_context, 1);
    if (context_create_return != GETDNS_RETURN_GOOD)
    {
        fprintf(stderr, "Trying to create the context failed: %d\n", context_create_return);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    /* Create an event base and put it in the context using the unknown function name */
    struct event_base *this_event_base;
    this_event_base = event_base_new();
    if (this_event_base == NULL)
    {
        fprintf(stderr, "Trying to create the event base failed.\n");
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    (void)getdns_extension_set_libevent_base(this_context, this_event_base);
    /* Set up the getdns call */
    const char * this_name  = "www.example.com";
    char* this_userarg = "somestring"; // Could add things here to help identify this call
    getdns_transaction_t this_transaction_id = 0;

    /* Make the call */
    getdns_return_t dns_request_return = getdns_address(this_context, this_name,
        NULL, this_userarg, &this_transaction_id, this_callbackfn);
    if (dns_request_return == GETDNS_RETURN_BAD_DOMAIN_NAME)
    {
        fprintf(stderr, "A bad domain name was used: %s. Exiting.\n", this_name);
        event_base_free(this_event_base);
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    else
    {
        /* Call the event loop */
        int dispatch_return = event_base_dispatch(this_event_base);
        UNUSED_PARAM(dispatch_return);
        // TODO: check the return value above
    }
    /* Clean up */
    event_base_free(this_event_base);
    getdns_context_destroy(this_context);
    /* Assuming we get here, leave gracefully */
    exit(EXIT_SUCCESS);
}

6.3 Get Addresses for a Domain Name And Their Associated DNSSEC Validation Status

This example shows how to check for secure DNSSEC results using the dnssec_return_status extension. In the innermost loop of the callback function, add a check for the DNSSEC status. It shows how to add two extensions to the extensions argument of the call.

getdns_dict * this_extensions = getdns_dict_create();
this_ret = getdns_dict_set_int(this_extensions, "return_both_v4_and_v6", GETDNS_EXTENSION_TRUE);
this_ret = getdns_dict_set_int(this_extensions, "dnssec_return_status", GETDNS_EXTENSION_TRUE);
. . .
if (this_type == GETDNS_RRTYPE_A)
{
    uint32_t this_dnssec_status;
    this_ret = getdns_dict_get_int(this_answer, "dnssec_status", &this_dnssec_status);
    if (this_dnssec_status != GETDNS_DNSSEC_SECURE)
    {
        // Log the DNSSEC status somewhere
    }
    else
    {
        // Deal with the record however you were going to
    }
}
. . .

You can put the DNSSEC status check outside the check for the particular type of record you care about, but you will then get log messages for bad status on records you might not care about as well.

6.4 Using the API Synchronously with getdns_general_sync()

This example is the same as the earlier examples, but uses getdns_general_sync() and thus does not use the async code. Note that the processing of the answers is essentially the same as it is for the synchronous example, it is just done in main().

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <getdns_core_only.h>

int main()
{
    getdns_return_t this_ret;  /* Holder for all function returns */
    /* Create the DNS context for this call */
    getdns_context *this_context = NULL;
    getdns_return_t context_create_return = getdns_context_create(&this_context, 1);
    if (context_create_return != GETDNS_RETURN_GOOD)
    {
        fprintf(stderr, "Trying to create the context failed: %d\n", context_create_return);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    /* Set up the getdns_sync_request call */
    const char * this_name  = "www.example.com";
    uint8_t this_request_type = GETDNS_RRTYPE_A;
    /* Get the A and AAAA records */
    getdns_dict * this_extensions = getdns_dict_create();
    this_ret = getdns_dict_set_int(this_extensions, "return_both_v4_and_v6", GETDNS_EXTENSION_TRUE);
    if (this_ret != GETDNS_RETURN_GOOD)
    {
        fprintf(stderr, "Trying to set an extension do both IPv4 and IPv6 failed: %d\n", this_ret);
        getdns_dict_destroy(this_extensions);
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    getdns_dict * this_response = NULL;

    /* Make the call */
    getdns_return_t dns_request_return = getdns_general_sync(this_context, this_name, this_request_type,
        this_extensions, &this_response);
    if (dns_request_return == GETDNS_RETURN_BAD_DOMAIN_NAME)
    {
        fprintf(stderr, "A bad domain name was used: %s. Exiting.\n", this_name);
        getdns_dict_destroy(this_response);
        getdns_dict_destroy(this_extensions);
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    else
    {
        /* Be sure the search returned something */
        uint32_t this_error;
        this_ret = getdns_dict_get_int(this_response, "status", &this_error);  // Ignore any error
        if (this_error != GETDNS_RESPSTATUS_GOOD)  // If the search didn't return "good"
        {
            fprintf(stderr, "The search had no results, and a return value of %d. Exiting.\n", this_error);
            getdns_dict_destroy(this_response);
            getdns_dict_destroy(this_extensions);
            getdns_context_destroy(this_context);
            return(GETDNS_RETURN_GENERIC_ERROR);
        }
        getdns_list * just_the_addresses_ptr;
        this_ret = getdns_dict_get_list(this_response, "just_address_answers", &just_the_addresses_ptr);  // Ignore any error
        size_t num_addresses;
        this_ret = getdns_list_get_length(just_the_addresses_ptr, &num_addresses);  // Ignore any error
        /* Go through each record */
        for ( size_t rec_count = 0; rec_count < num_addresses; ++rec_count )
        {
            getdns_dict * this_address;
            this_ret = getdns_list_get_dict(just_the_addresses_ptr, rec_count, &this_address);  // Ignore any error
            /* Just print the address */
            getdns_bindata * this_address_data;
            this_ret = getdns_dict_get_bindata(this_address, "address_data", &this_address_data); // Ignore any error
            char *this_address_str = getdns_display_ip_address(this_address_data);
            printf("The address is %s\n", this_address_str);
            free(this_address_str);
        }
    }
    /* Clean up */
    getdns_dict_destroy(this_response); 
    getdns_dict_destroy(this_extensions);
    getdns_context_destroy(this_context);
    /* Assuming we get here, leave gracefully */
    exit(EXIT_SUCCESS);
}

6.5 Getting Names from the Reverse Tree with getdns_hostname()

This example shows how to use getdns_hostname() to get names from the DNS reverse tree.

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <getdns_libevent.h>

#define UNUSED_PARAM(x) ((void)(x))

/* Set up the callback function, which will also do the processing of the results */
void this_callbackfn(getdns_context *this_context,
                     getdns_callback_type_t this_callback_type,
                     getdns_dict *this_response, 
                     void *this_userarg,
                     getdns_transaction_t this_transaction_id)
{
    getdns_return_t this_ret;  /* Holder for all function returns */
    UNUSED_PARAM(this_userarg);  /* Not looking at the userarg for this example */
    UNUSED_PARAM(this_context);  /* Not looking at the context for this example */
    UNUSED_PARAM(this_ret); /* Set, but not read */
    if (this_callback_type == GETDNS_CALLBACK_COMPLETE)  /* This is a callback with data */
    {
        /* Be sure the search returned something */
        uint32_t this_error;
        this_ret = getdns_dict_get_int(this_response, "status", &this_error);  // Ignore any error
        if (this_error != GETDNS_RESPSTATUS_GOOD)  // If the search didn't return "good"
        {
            fprintf(stderr, "The search had no results, and a return value of %d. Exiting.\n", this_error);
            getdns_dict_destroy(this_response);
            return;
        }
        getdns_list *replies_tree;
        this_ret = getdns_dict_get_list(this_response, "replies_tree", &replies_tree);  // Ignore any error
        size_t num_replies;
        this_ret = getdns_list_get_length(replies_tree, &num_replies);  // Ignore any error
        /* Go through each reply */
        for ( size_t reply_count = 0; reply_count < num_replies; ++reply_count)
        {
            getdns_dict * this_reply;
            this_ret = getdns_list_get_dict(replies_tree, reply_count, &this_reply);  // Ignore any error
            /* Just print the address */
            getdns_list* reply_answers;
            this_ret = getdns_dict_get_list(this_reply, "answer", &reply_answers); // Ignore any error
            size_t num_answers;
            this_ret = getdns_list_get_length(reply_answers, &num_answers);  // Ignore any error
            /* Go through each answer */
            for ( size_t answer_count = 0; answer_count < num_answers; ++answer_count)
            {
                getdns_dict * this_rr;
                this_ret = getdns_list_get_dict(reply_answers, answer_count, &this_rr);
                /* Get the RDATA type */
                uint32_t this_type;
                this_ret = getdns_dict_get_int(this_rr, "type", &this_type);  // Ignore any error
                if (this_type == GETDNS_RRTYPE_PTR)
                {
                    getdns_dict *this_rdata;
                    this_ret = getdns_dict_get_dict(this_rr, "rdata", &this_rdata);  // Ignore any error

                    getdns_bindata * this_dname;
                    this_ret = getdns_dict_get_bindata(this_rdata, "rdata_raw", &this_dname);
                    char *this_dname_str;
                           this_ret = getdns_convert_dns_name_to_fqdn(this_dname, &this_dname_str); // Ignore any error
                    printf("The dname is %s\n", this_dname_str);
                    free(this_dname_str);
                }
            }

        }
    }
    else if (this_callback_type == GETDNS_CALLBACK_CANCEL)
        fprintf(stderr, "The callback with ID %"PRIu64" was cancelled. Exiting.", this_transaction_id);
    else
        fprintf(stderr, "The callback got a callback_type of %d. Exiting.", this_callback_type);
    getdns_dict_destroy(this_response);
}

int main()
{
    /* Create the DNS context for this call */
    getdns_context *this_context = NULL;
    getdns_return_t context_create_return = getdns_context_create(&this_context, 1);
    if (context_create_return != GETDNS_RETURN_GOOD)
    {
        fprintf(stderr, "Trying to create the context failed: %d", context_create_return);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    /* Create an event base and put it in the context using the unknown function name */
    struct event_base *this_event_base;
    this_event_base = event_base_new();
    if (this_event_base == NULL)
    {
        fprintf(stderr, "Trying to create the event base failed.");
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    (void)getdns_extension_set_libevent_base(this_context, this_event_base);
    /* Set up the getdns call */
    getdns_dict * this_addr_to_look_up = getdns_dict_create();
    // TODO: check the return value above
    getdns_bindata this_type = { 4, (void *)"IPv4" };
    getdns_return_t this_ret = getdns_dict_set_bindata(this_addr_to_look_up, "address_type", &this_type);
    UNUSED_PARAM(this_ret);
    getdns_bindata this_ipv4_addr = { 4, (void *)"\x08\x08\x08\x08" };
    this_ret = getdns_dict_set_bindata(this_addr_to_look_up, "address_data", &this_ipv4_addr);
    char* this_userarg = "somestring"; // Could add things here to help identify this call
    getdns_transaction_t this_transaction_id = 0;

    /* Make the call */
    getdns_return_t dns_request_return = getdns_hostname(this_context, this_addr_to_look_up,
        NULL, this_userarg, &this_transaction_id, this_callbackfn);
    if (dns_request_return == GETDNS_RETURN_BAD_DOMAIN_NAME)
    {
        char *ip_address_str = getdns_display_ip_address(&this_ipv4_addr);

        fprintf(stderr, "A bad IP address was used: %s. Exiting.\n", ip_address_str);
        free(ip_address_str);
        getdns_dict_destroy(this_addr_to_look_up);
        event_base_free(this_event_base);
        getdns_context_destroy(this_context);
        return(GETDNS_RETURN_GENERIC_ERROR);
    }
    else
    {
        /* Call the event loop */
        int dispatch_return = event_base_dispatch(this_event_base);
        UNUSED_PARAM(dispatch_return);
        // TODO: check the return value above
    }
    /* Clean up */
    getdns_dict_destroy(this_addr_to_look_up);
    event_base_free(this_event_base);
    getdns_context_destroy(this_context);
    /* Assuming we get here, leave gracefully */
    exit(EXIT_SUCCESS);
}

7. More Helper Functions

The following two functions convert individual labels of IDNs between their Unicode encoding and their ASCII encoding. They follow the rules for IDNA 2008 described in RFC 5890-5892.

If an application wants the API do perform DNSSEC validation without using the extensions, it can use the getdns_validate_dnssec() helper function.

The record_to_validate is the resource record being validated together with the associated signatures. The API will use the resource records in bundle_of_support_records to construct the validation chain and the DNSKEY or DS records in trust_anchor_records as trust anchors. The function returns one of GETDNS_DNSSEC_SECURE, GETDNS_DNSSEC_BOGUS, GETDNS_DNSSEC_INDETERMINATE, or GETDNS_DNSSEC_INSECURE.

The default list of trust anchor records that is used by the library to validate DNSSEC can be retrieved by using the getdns_root_trust_anchor helper function.

When there are no default trust anchors NULL is returned. Upon successful return, the variable of type time_t, referenced by utc_date_of_anchor is set to the number of seconds since epoch the trust anchors were obtained.

There are two functions that help process data:

This returns a string that is the nicely-formatted version of the dict and all of the named elements in it.

This returns a string that is the nicely-formatted version of the IPv4 or IPv6 address in it. The API determines they type of address by the length given in the bindata.

All memory locations returned by these helper functions are allocated by the default system allocator, namely malloc. The caller is responsible of disposing these allocations with free.

8. DNS Contexts

Many calls in the DNS API require a DNS context. A DNS context contains the information that the API needs in order to process DNS calls, such as the locations of upstream DNS servers, DNSSEC trust anchors, and so on. The internal structure of the DNS context is opaque, and might be different on each OS. When a context is passed to any function, it must be an allocated context; the context must not be NULL.

A typical application using this API doesn't need to know anything about contexts. Basically, the application creates a default context, uses it in the functions that require a context, and then deallocates it when done. Context manipulation is available for more DNS-aware programs, but is unlikely to be of interest to applications that just want the results of lookups for A, AAAA, SRV, and PTR records.

It is expected that contexts in implementations of the API will not necessarily be thread-safe, but they will not be thread-hostile. A context should not be used by multiple threads: create a new context for use on a different thread. It is just fine for an application to have many contexts, and some DNS-heavy applications will certainly want to have many even if the application uses a single thread.

See above for the method for creating and destroying contexts. When the context is used in the API for the first time and set_from_os is 1, the API starts replacing some of the values with values from the OS, such as those that would be found in res_query(3), /etc/resolv.conf, and so on, then proceeds with the new function. Some advanced users will not want the API to change the values to the OS's defaults; if set_from_os is 0, the API will not do any updates to the initial values based on changes in the OS. For example, this might be useful if the API is acting as a stub resolver that is using a specific upstream recursive resolver chosen by the application, not the one that might come back from DHCP.

8.1 Updating the Context Automatically

The context returned by getdns_context_create() is updated by the API by default, such as when changes are made to /etc/resolv.conf. When there is a change, the callback function that is set in getdns_context_set_context_update_callback() (described below) is called.

Many of the defaults for a context come from the operating system under which the API is running. In specific, it is important that the implementation should try to replicate as best as possible the logic of a local getaddrinfo() when creating a new context. This includes making lookups in WINS for NetBIOS, mDNS lookups, nis names, and any other name lookup that getaddrinfo() normally does automatically. The API should look at nsswitch, the Windows resolver, and so on.

In the function definitions below, the choice listed in bold is the one used for the API default context.

8.2 Updating the Context Manually

Setting specific values in a context are done with value-specific functions shown here. The setting functions all return either GETDNS_RETURN_GOOD for success or GETDNS_RETURN_CONTEXT_UPDATE_FAIL for a failure to update the context.

An application can be notified when the context is changed.

The value is a pointer to the callback function that will be called when any context is changed. Such changes might be from automatic changes from the API (such as changes to /etc/resolv.conf), or might be from any of the API functions in this section being called. The second argument to the callback function specifies which of the context changed; the context codes are listed later in this document.

Calling getdns_context_set_context_update_callback with a second argument of NULL prevents updates to the context from causing callbacks.

8.3 Contexts for Basic Resolution

Specifies whether DNS queries are performed with nonrecurive lookups or as a stub resolver. The value is GETDNS_RESOLUTION_RECURSING or GETDNS_RESOLUTION_STUB.

All implementations of this API can act as recursive resolvers, and that must be the default mode of the default context. Some implementations of this API are expected to also be able to act as stub resolvers. If an implementation of this API is only able to act as a recursive resolver, a call to getdns_context_set_resolution_type(somecontext, GETDNS_RESOLUTION_STUB) will return GETDNS_RETURN_CONTEXT_UPDATE_FAIL.

The namespaces array contains an ordered list of namespaces that will be queried. Important: this context setting is ignored for the getdns_general and getdns_general_sync functions; it is used for the other funtions. The values are GETDNS_NAMESPACE_DNS, GETDNS_NAMESPACE_LOCALNAMES, GETDNS_NAMESPACE_NETBIOS, GETDNS_NAMESPACE_MDNS, and GETDNS_NAMESPACE_NIS. When a normal lookup is done, the API does the lookups in the order given and stops when it gets the first result; a different method with the same result would be to run the queries in parallel and return when it gets the first result. Because lookups might be done over different mechanisms because of the different namespaces, there can be information leakage that is similar to that seen with getaddrinfo(). The default is determined by the OS.

Specifies what transport is used for DNS lookups. The value is GETDNS_TRANSPORT_UDP_FIRST_AND_FALL_BACK_TO_TCP, GETDNS_TRANSPORT_UDP_ONLY, GETDNS_TRANSPORT_TCP_ONLY, or GETDNS_TRANSPORT_TCP_ONLY_KEEP_CONNECTIONS_OPEN.

Specifies limit the number of outstanding DNS queries. The API will block itself from sending more queries if it is about to exceed this value, and instead keep those queries in an internal queue. The a value of 0 indicates that the number of outstanding DNS queries is unlimited.

Specifies number of milliseconds the API will wait for request to return. The default is not specified.

8.4 Context for Recursive Resolvers

Specifies whether or not DNS queries follow redirects. The value is GETDNS_REDIRECTS_FOLLOW for normal following of redirects though CNAME and DNAME; or GETDNS_REDIRECTS_DO_NOT_FOLLOW to cause any lookups that would have gone through CNAME and DNAME to return the CNAME or DNAME, not the eventual target.

The list contains dicts that are addresses to be used for looking up top-level domains; the default is the list of "normal" IANA root servers. Each dict in the list contains at least two names: address_type (whose value is a bindata; it is currently either "IPv4" or "IPv6") and address_data (whose value is a bindata).

8.5 Context for Local Naming

Specifies whether to append a suffix to the query string before the API starts resolving a name. The value is GETDNS_APPEND_NAME_ALWAYS, GETDNS_APPEND_NAME_ONLY_TO_SINGLE_LABEL_AFTER_FAILURE, GETDNS_APPEND_NAME_ONLY_TO_MULTIPLE_LABEL_NAME_AFTER_FAILURE, or GETDNS_APPEND_NAME_NEVER. This controls whether or not to append the suffix given by getdns_context_set_suffix

The value is a list of bindatas that are strings that are to be appended based on getdns_context_set_append_name; the default is an empty list. The values here follow the rules in section 2.1 of RFC 4343 to allow non-ASCII octets and special characters in labels.

8.6 Context for DNSSEC

These context settings affect queries that have extensions that specify the use of DNSSEC.

Applications that need to specify the DNSSEC trust anchors can use:

The value is a list of bindatas that are the DNSSEC trust anchors. The default is the trust anchors from the IANA root. The trust anchors are expressed as RDATAs from DNSKEY resource records.

In the rare case that an application needs to set the DNSSEC skew, it can:

The value is the number of seconds of skew that is allowed in either direction when checking an RRSIG's Expiration and Inception fields. The default is 0.

8.7 Context Specific to Stub Resolvers

An application can change the quering mechanism of a context to be to act as a stub resolver. Such an application might first get the default information to make this change from the operating system, probably through DHCP.

Note that if a context is changed to being a stub resolver, this automatically prevents the application from using the extenstions for DNSSEC. An application that wants to both do DNSSEC and stub resolution must do its own DNSSEC processing, possibly with the getdns_validate_dnssec() function.

The list of dicts define where a stub resolver will send queries. Each dict contains at least two names: address_type (whose value is a bindata; it is currently either "IPv4" or "IPv6") and address_data (whose value is a bindata). It might also contain port to specify which port to use to contact these DNS servers; the default is 53. If the stub and a recursive resolver both support TSIG (RFC 2845), the upstream_list entry can also contain tsig_algorithm (a bindata) that is the name of the TSIG hash algorithm, and tsig_secret (a bindata) that is the TSIG key.

8.8 Context for EDNS

These context settings affect queries that have extensions that specify the use of OPT resource records. These come from RFC 6891.

The value is between 512 and 65535; the default is 512.

The value is between 0 and 255; the default is 0.

The value is between 0 and 255; the default is 0.

The value is between 0 and 1; the default is 0.

8.9 Context Use of Custom Memory Management Functions

The given memory management functions will be used for creating the response dicts. The response dicts inherit the custom memory management functions from the context and will deallocate themselves (and their members) with the custom deallocator. By default, the system malloc, realloc, and free are used.

The given extended memory management functions will be used for creating the response dicts. The value of userarg argument will be passed to the custom malloc, realloc, and free. The response dicts inherit the custom memory management functions and the value for userarg from the context and will deallocate themselves (and their members) with the custom deallocator.

8.10 Context Codes

The context codes for getdns_context_set_context_update_callback() are:

GETDNS_CONTEXT_CODE_NAMESPACES

Change related to getdns_context_set_namespaces

GETDNS_CONTEXT_CODE_RESOLUTION_TYPE

Change related to getdns_context_set_resolution_type

GETDNS_CONTEXT_CODE_FOLLOW_REDIRECTS

Change related to getdns_context_set_follow_redirects

GETDNS_CONTEXT_CODE_UPSTREAM_RECURSIVE_SERVERS

Change related to getdns_context_set_upstream_recursive_servers

GETDNS_CONTEXT_CODE_DNS_ROOT_SERVERS

Change related to getdns_context_set_dns_root_servers

GETDNS_CONTEXT_CODE_DNS_TRANSPORT

Change related to getdns_context_set_dns_transport

GETDNS_CONTEXT_CODE_LIMIT_OUTSTANDING_QUERIES

Change related to getdns_context_set_limit_outstanding_queries

GETDNS_CONTEXT_CODE_APPEND_NAME

Change related to getdns_context_set_append_name

GETDNS_CONTEXT_CODE_SUFFIX

Change related to getdns_context_set_suffix

GETDNS_CONTEXT_CODE_DNSSEC_TRUST_ANCHORS

Change related to getdns_context_set_dnssec_trust_anchors

GETDNS_CONTEXT_CODE_EDNS_MAXIMUM_UDP_PAYLOAD_SIZE

Change related to getdns_context_set_edns_maximum_udp_payload_size

GETDNS_CONTEXT_CODE_EDNS_EXTENDED_RCODE

Change related to getdns_context_set_edns_extended_rcode

GETDNS_CONTEXT_CODE_EDNS_VERSION

Change related to getdns_context_set_edns_version

GETDNS_CONTEXT_CODE_EDNS_DO_BIT

Change related to getdns_context_set_edns_do_bit

GETDNS_CONTEXT_CODE_DNSSEC_ALLOWED_SKEW

Change related to getdns_context_set_dnssec_allowed_skew

GETDNS_CONTEXT_CODE_MEMORY_FUNCTIONS

Change related to getdns_context_set_memory_functions

GETDNS_CONTEXT_CODE_TIMEOUT

Change related to getdns_context_set_timeout

8.11 Getting API information and current Contexts

An application might want to see information about the API itself and inspect the current context. Use the getdns_context_get_api_information function.

The returned getdns_dict will contain the following name/value pairs:

• version_string (a bindata) represents the version string for this version of the DNS API.
• implementation_string (a bindata) is a string set by the API implementer. It might be human-readable, and it might have information in it useful to an application developer (but it doesn't have to).
• resolution_type (an int) is the type of resolver that the API is acting as in this context: GETDNS_RESOLUTION_RECURSING or GETDNS_RESOLUTION_STUB (it will be a recursing resolver unless the application changed this in a context.
• all_context (a dict) with names for all the types of context. This can be used with getdns_pretty_print_dict() for debugging.

9. The Generated Files

There is a tarball that includes the .h files, the examples, and so on. The examples all make, even though there is no API implementation, based on a pseudo-implementation in the tarball; see make-examples-PLATFORM.sh. Note that this currently builds fine on the Macintosh and Ubuntu; help is definitely appreciated on making the build process work on more platforms if it fails there.

10. Commentary

The following description of the API may be of value to those who might implement the design, and those who are using an implementation of the design.

10.1 API Design Considerations

The genesis of this DNS API design was seeing other DNS API designs flounder. There are other DNS APIs already available (such as draft-hayatnagarkar-dnsext-validator-api, as well as DNSSEC APIs in BIND and Unbound), but there has been very little uptake of them. In talking to application developers, there was a consistent story: that they felt that the APIs were developed by and for DNS people, not applications developers.

This API design comes from talking to a small handful of applications developers about what they would want to see in a modern DNS API. Now that the API is public, it would be great to hear from many more application developers about whether it would meet their needs if it was implemented. My goal is to create a design that is a natural follow-on to getaddrinfo() that has all the capabilities that most application developers might want now or in the next few years: access to all types of DNS records (including those which are yet to be defined), full DNSSEC awareness, IDN handling, and parity for IPv4 and IPv6 addresses.

Note that this is just a design for a new API: there is no implementation of the design yet, but at least one is being worked on. The process of designing the API without implementing it at the same time has the huge advantage that major design changes could be made without any worry about "but we already coded it the other way". In the early revisions of this document, many fundamental design choices changed over and over, and even bike-shedding-like changes were allowed because they didn't involve any programming effort.

This work was done independently, not through the IETF because the IETF generally doesn't take on API work, and has explicitly avoided DNS API work in the past.

This API design has a Creative Commons license so that it can be used widely by potential API implementers. This also allows other people who want to fork the design to do so cleanly. Of course, any implementation of this API can choose whatever kind of license the API implementer wishes, but it would be fabulous if one or more such implementations had Creative Commons or BSD-ish licenses.

The API relies heavily on C macros and hopefully has no magic numbers.

10.2 API Implementation Considerations

All implementations of this API must act as recursive resolvers, and some might choose not to be able to act as stub resolvers. Note that all implementations of this API must be DNSSEC validators.

Because there are many C event libraries available, and they have different calling routines, it is the implementation of an API that determines which event library is used. This is certainly not optimal for C programmers, but they appear to have gotten used to is so far. All implementations of this API must support synchronous calls with getdns_general_sync().

Versions are differentiated by version strings instead of version numbers. The version string for this API is "getdns April 2013". Each implementation is free to set the implementation string as it feels fit.

The API's .h file contains a macro called GETDNS_COMPILATION_COMMENT. This can be useful to an application which will use the API because it can check the string without calling any functions. Each time the API implementation is compiled, this string should be updated with unique information about the implementation build.

The implementation of both the async and sync getdns functions will copy all the values of the parameters into local memory, in case the application changes or deallocates them.

This work is licensed under a Creative Commons Attribution 3.0 Unported License.