Methods¶

ResolveHostname() takes a hostname and resolves it to one or more IP addresses. As parameters it takes the Linux network interface index to execute the query on, or 0 if it may be done on any suitable interface. The name parameter specifies the hostname to resolve. Note that if required, IDNA conversion is applied to this name unless it is resolved via LLMNR or MulticastDNS. The family parameter limits the results to a specific address family. It may be AF_INET, AF_INET6 or AF_UNSPEC. If AF_UNSPEC is specified (recommended), both kinds are retrieved, subject to local network configuration (i.e. if no local, routable IPv6 address is found, no IPv6 address is retrieved; and similarly for IPv4). A 64-bit flags field may be used to alter the behaviour of the resolver operation (see below). The method returns an array of address records. Each address record consists of the interface index the address belongs to, an address family as well as a byte array with the actual IP address data (which either has 4 or 16 elements, depending on the address family). The returned address family will be one of AF_INET or AF_INET6. For IPv6, the returned address interface index should be used to initialize the .sin6_scope_id field of a struct sockaddr_in6 instance to permit support for resolution to link-local IP addresses. The address array is followed by the canonical name of the host, which may or may not be identical to the resolved hostname. Finally, a 64-bit flags field is returned that is defined similarly to the flags field that was passed in, but contains information about the resolved data (see below). If the hostname passed in is an IPv4 or IPv6 address formatted as string, it is parsed, and the result is returned. In this case, no network communication is done.

ResolveAddress() executes the reverse operation: it takes an IP address and acquires one or more hostnames for it. As parameters it takes the interface index to execute the query on, or 0 if all suitable interfaces are OK. The family parameter indicates the address family of the IP address to resolve. It may be either AF_INET or AF_INET6. The address parameter takes the raw IP address data (as either a 4 or 16 byte array). The flags input parameter may be used to alter the resolver operation (see below). The method returns an array of name records, each consisting of an interface index and a hostname. The flags output field contains additional information about the resolver operation (see below).

ResolveRecord() takes a DNS resource record (RR) type, class and name, and retrieves the full resource record set (RRset), including the RDATA, for it. As parameter it takes the Linux network interface index to execute the query on, or 0 if it may be done on any suitable interface. The name parameter specifies the RR domain name to look up (no IDNA conversion is applied), followed by the 16-bit class and type fields (which may be ANY). Finally, a flags field may be passed in to alter behaviour of the look-up (see below). On completion, an array of RR items is returned. Each array entry consists of the network interface index the RR was discovered on, the type and class field of the RR found, and a byte array of the raw RR discovered. The raw RR data starts with the RR's domain name, in the original casing, followed by the RR type, class, TTL and RDATA, in the binary format documented in RFC 1035. For RRs that support name compression in the payload (such as MX or PTR), the compression is expanded in the returned data.

Note that currently, the class field has to be specified as IN or ANY. Specifying a different class will return an error indicating that look-ups of this kind are unsupported. Similarly, some special types are not supported either (AXFR, OPT, …). While systemd-resolved parses and validates resource records of many types, it is crucial that clients using this API understand that the RR data originates from the network and should be thoroughly validated before use.

ResolveService() may be used to resolve a DNS SRV service record, as well as the hostnames referenced in it, and possibly an accompanying DNS-SD TXT record containing additional service metadata. The primary benefit of using this method over ResolveRecord() specifying the SRV type is that it will resolve the SRV and TXT RRs as well as the hostnames referenced in the SRV in a single operation. As parameters it takes a Linux network interface index, a service name, a service type and a service domain. This method may be invoked in three different modes:

The family parameter of the ResolveService() method encodes the desired family of the addresses to resolve (use AF_INET, AF_INET6, or AF_UNSPEC). If this is enabled (Use the NO_ADDRESS flag to turn address resolution off, see below). The flags parameter takes a couple of flags that may be used to alter the resolver operation.

On completion, ResolveService() returns an array of SRV record structures. Each items consisting of the priority, weight and port fields as well as the hostname to contact, as encoded in the SRV record. Immediately following is an array of the addresses of this hostname, with each item consisting of the interface index, the address family and the address data in a byte array. This address array is followed by the canonicalized hostname. After this array of SRV record structures an array of byte arrays follows that encodes the TXT RR strings, in case DNS-SD look-ups are enabled. The next parameters are the canonical service name, type and domain. This may or may not be identical to the parameters passed in. Finally, a flags field is returned that contains information about the resolver operation performed.

The ResetStatistics() method resets the various statistics counters that systemd-resolved maintains to zero. (For details, see the statistics properties below.)

The GetLink() method takes a network interface index and returns the object path to the org.freedesktop.resolve1.Link object corresponding to it.

The SetLinkDNS() method sets the DNS servers to use on a specific interface. This method (and the following ones) may be used by network management software to configure per-interface DNS settings. It takes a network interface index as well as an array of DNS server IP address records. Each array item consists of an address family (either AF_INET or AF_INET6), followed by a 4-byte or 16-byte array with the raw address data. This method is a one-step shortcut for retrieving the Link object for a network interface using GetLink() (see above) and then invoking the SetDNS() method (see below) on it.

SetLinkDNSEx() is similar to SetLinkDNS(), but allows an IP port (instead of the default 53) and DNS name to be specified for each DNS server. The server name is used for Server Name Indication (SNI), which is useful when DNS-over-TLS is used. C.f. DNS= in resolved.conf(5).

SetLinkDefaultRoute() specifies whether the link shall be used as the default route for name queries. See the description of name routing in systemd-resolved.service(8) for details.

The SetLinkDomains() method sets the search and routing domains to use on a specific network interface for DNS look-ups. It takes a network interface index and an array of domains, each with a boolean parameter indicating whether the specified domain shall be used as a search domain (false), or just as a routing domain (true). Search domains are used for qualifying single-label names into FQDN when looking up hostnames, as well as for making routing decisions on which interface to send queries ending in the domain to. Routing domains are only used for routing decisions and not used for single-label name qualification. Pass the search domains in the order they should be used.

The SetLinkLLMNR() method enables or disables LLMNR support on a specific network interface. It takes a network interface index as well as a string that may either be empty or one of "yes", "no" or "resolve". If empty, the systemd-wide default LLMNR setting is used. If "yes", LLMNR is used for resolution of single-label names and the local hostname is registered on all local LANs for LLMNR resolution by peers. If "no", LLMNR is turned off fully on this interface. If "resolve", LLMNR is only enabled for resolving names, but the local hostname is not registered for other peers to use.

Similarly, the SetLinkMulticastDNS() method enables or disables MulticastDNS support on a specific interface. It takes the same parameters as SetLinkLLMNR() described above.

The SetLinkDNSSEC() method enables or disables DNSSEC validation on a specific network interface. It takes a network interface index as well as a string that may either be empty or one of "yes", "no", or "allow-downgrade". When empty, the system-wide default DNSSEC setting is used. If "yes", full DNSSEC validation is done for all look-ups. If the selected DNS server does not support DNSSEC, look-ups will fail if this mode is used. If "no", DNSSEC validation is fully disabled. If "allow-downgrade", DNSSEC validation is enabled, but is turned off automatically if the selected server does not support it (thus opening up behaviour to downgrade attacks). Note that DNSSEC only applies to traditional DNS, not to LLMNR or MulticastDNS.

The SetLinkDNSSECNegativeTrustAnchors() method may be used to configure DNSSEC Negative Trust Anchors (NTAs) for a specific network interface. It takes a network interface index and a list of domains as arguments.

The SetLinkDNSOverTLS() method enables or disables DNS-over-TLS. C.f. DNSOverTLS= in systemd-resolved.service(8) for details.

Network management software integrating with systemd-resolved should call SetLinkDNS() or SetLinkDNSEx(), SetLinkDefaultRoute(), SetLinkDomains() and others after the interface appeared in the kernel (and thus after a network interface index has been assigned), but before the network interfaces is activated (IFF_UP set) so that all settings take effect during the full time the network interface is up. It is safe to alter settings while the interface is up, however. Use RevertLink() (described below) to reset all per-interface settings.

The RevertLink() method may be used to revert all per-link settings described above to the defaults.

The FlushCaches() flushes all resource record caches maintained by the resolver, and ensures that any subsequent lookups re-request their responses from their sources.

The ResetServerFeatures() flushes any feature information learned about remote DNS servers. This ensures that subsequent lookups will be initially attempted at the highest DNS protocol feature level again, possibly requiring a (potentially slow) downgrade cycle to recognize the supported feature level again.

The RegisterService() method may be used to register a DNS-SD service on the host. This functionality is closely related to the functionality provided by systemd.dnssd(5) files. It takes a server identifier string as first parameter (this is just a local identifier, and should be chosen so that it neither collides with the basename of *.dnssd files nor with names chosen by other IPC clients). It also takes a name template string for the DNS-SD service name visible on the network. This string is subject to specifier expansation, as documented for the Name= setting in *.dnssd files. It also takes a service type string containing the DNS-SD service type, as well as an IP port, a priority/weight pair for the DNS-SD SRV record. Finally, it takes an array of TXT record data. It returns an object path which may be used as handle to the registered service.

The UnregisterService() method undoes the effect of RegisterService() and deletes a DNS-SD service previously created via IPC again.

/* Input+Output: Protocol/scope */
#define SD_RESOLVED_DNS                (UINT64_C(1) <<  0)
#define SD_RESOLVED_LLMNR_IPV4         (UINT64_C(1) <<  1)
#define SD_RESOLVED_LLMNR_IPV6         (UINT64_C(1) <<  2)
#define SD_RESOLVED_MDNS_IPV4          (UINT64_C(1) <<  3)
#define SD_RESOLVED_MDNS_IPV6          (UINT64_C(1) <<  4)

/* Input: Restrictions */
#define SD_RESOLVED_NO_CNAME           (UINT64_C(1) <<  5)
#define SD_RESOLVED_NO_TXT             (UINT64_C(1) <<  6)
#define SD_RESOLVED_NO_ADDRESS         (UINT64_C(1) <<  7)
#define SD_RESOLVED_NO_SEARCH          (UINT64_C(1) <<  8)
#define SD_RESOLVED_NO_VALIDATE        (UINT64_C(1) << 10)
#define SD_RESOLVED_NO_SYNTHESIZE      (UINT64_C(1) << 11)
#define SD_RESOLVED_NO_CACHE           (UINT64_C(1) << 12)
#define SD_RESOLVED_NO_ZONE            (UINT64_C(1) << 13)
#define SD_RESOLVED_NO_TRUST_ANCHOR    (UINT64_C(1) << 14)
#define SD_RESOLVED_NO_NETWORK         (UINT64_C(1) << 15)
#define SD_RESOLVED_NO_STALE           (UINT64_C(1) << 24)
#define SD_RESOLVED_RELAX_SINGLE_LABEL (UINT64_C(1) << 25)

/* Output: Security */
#define SD_RESOLVED_AUTHENTICATED      (UINT64_C(1) <<  9)
#define SD_RESOLVED_CONFIDENTIAL       (UINT64_C(1) << 18)

/* Output: Origin */
#define SD_RESOLVED_SYNTHETIC          (UINT64_C(1) << 19)
#define SD_RESOLVED_FROM_CACHE         (UINT64_C(1) << 20)
#define SD_RESOLVED_FROM_ZONE          (UINT64_C(1) << 21)
#define SD_RESOLVED_FROM_TRUST_ANCHOR  (UINT64_C(1) << 22)
#define SD_RESOLVED_FROM_NETWORK       (UINT64_C(1) << 23)

Properties¶

The LLMNR and MulticastDNS properties report whether LLMNR and MulticastDNS are (globally) enabled. Each may be one of "yes", "no", and "resolve". See SetLinkLLMNR() and SetLinkMulticastDNS() above.

LLMNRHostname contains the hostname currently exposed on the network via LLMNR. It usually follows the system hostname as may be queried via gethostname(3), but may differ if a conflict is detected on the network.

DNS and DNSEx contain arrays of all DNS servers currently used by systemd-resolved. DNS contains information similar to the DNS server data in /run/systemd/resolve/resolv.conf. Each structure in the array consists of a numeric network interface index, an address family, and a byte array containing the DNS server address (either 4 bytes in length for IPv4 or 16 bytes in lengths for IPv6). DNSEx is similar, but additionally contains the IP port and server name (used for Server Name Indication, SNI). Both arrays contain DNS servers configured system-wide, including those possibly read from a foreign /etc/resolv.conf or the DNS= setting in /etc/systemd/resolved.conf, as well as per-interface DNS server information either retrieved from systemd-networkd(8), or configured by external software via SetLinkDNS() or SetLinkDNSEx() (see above). The network interface index will be 0 for the system-wide configured services and non-zero for the per-link servers.

FallbackDNS and FallbackDNSEx contain arrays of all DNS servers configured as fallback servers, if any, using the same format as DNS and DNSEx described above. See the description of FallbackDNS= in resolved.conf(5) for the description of when those servers are used.

CurrentDNSServer and CurrentDNSServerEx specify the server that is currently used for query resolution, in the same format as a single entry in the DNS and DNSEx arrays described above.

Similarly, the Domains property contains an array of all search and routing domains currently used by systemd-resolved. Each entry consists of a network interface index (again, 0 encodes system-wide entries), the actual domain name, and whether the entry is used only for routing (true) or for both routing and searching (false).

The TransactionStatistics property contains information about the number of transactions systemd-resolved has processed. It contains a pair of unsigned 64-bit counters, the first containing the number of currently ongoing transactions, the second the number of total transactions systemd-resolved is processing or has processed. The latter value may be reset using the ResetStatistics() method described above. Note that the number of transactions does not directly map to the number of issued resolver bus method calls. While simple look-ups usually require a single transaction only, more complex look-ups might result in more, for example when CNAMEs or DNSSEC are in use.

The CacheStatistics property contains information about the executed cache operations so far. It exposes three 64-bit counters: the first being the total number of current cache entries (both positive and negative), the second the number of cache hits, and the third the number of cache misses. The latter counters may be reset using ResetStatistics() (see above).

The DNSSEC property specifies current status of DNSSEC validation. It is one of "yes" (validation is enforced), "no" (no validation is done), "allow-downgrade" (validation is done if the current DNS server supports it). See the description of DNSSEC= in resolved.conf(5).

The DNSSECStatistics property contains information about the DNSSEC validations executed so far. It contains four 64-bit counters: the number of secure, insecure, bogus, and indeterminate DNSSEC validations so far. The counters are increased for each validated RRset, and each non-existence proof. The secure counter is increased for each operation that successfully verified a signed reply, the insecure counter is increased for each operation that successfully verified that an unsigned reply is rightfully unsigned. The bogus counter is increased for each operation where the validation did not check out and the data is likely to have been tempered with. Finally the indeterminate counter is increased for each operation which did not complete because the necessary keys could not be acquired or the cryptographic algorithms were unknown.

The DNSSECSupported boolean property reports whether DNSSEC is enabled and the selected DNS servers support it. It combines information about system-wide and per-link DNS settings (see below), and only reports true if DNSSEC is enabled and supported on every interface for which DNS is configured and for the system-wide settings if there are any. Note that systemd-resolved assumes DNSSEC is supported by DNS servers until it verifies that this is not the case. Thus, the reported value may initially be true, until the first transactions are executed.

The DNSOverTLS boolean property reports whether DNS-over-TLS is enabled.

The ResolvConfMode property exposes how /etc/resolv.conf is managed on the host. Currently, the values "uplink", "stub", "static" (these three correspond to the three different files systemd-resolved.service provides), "foreign" (the file is managed by admin or another service, systemd-resolved.service just consumes it), "missing" (/etc/resolv.conf is missing).

The DNSStubListener property reports whether the stub listener on port 53 is enabled. Possible values are "yes" (enabled), "no" (disabled), "udp" (only the UDP listener is enabled), and "tcp" (only the TCP listener is enabled).

The DNSSECNegativeTrustAnchors property contains a list of recognized DNSSEC negative trust anchors and contains a list of domains.

Methods¶

The various methods exposed by the Link interface are equivalent to their similarly named counterparts on the Manager interface. e.g. SetDNS() on the Link object maps to SetLinkDNS() on the Manager object, the main difference being that the later expects an interface index to be specified. Invoking the methods on the Manager interface has the benefit of reducing roundtrips, as it is not necessary to first request the Link object path via GetLink() before invoking the methods. The same relationship holds for SetDNSEx(), SetDomains(), SetDefaultRoute(), SetLLMNR(), SetMulticastDNS(), SetDNSOverTLS(), SetDNSSEC(), SetDNSSECNegativeTrustAnchors(), and Revert(). For further details on these methods see the Manager documentation above.

Properties¶

ScopesMask defines which resolver scopes are currently active on this interface. This 64-bit unsigned integer field is a bit mask consisting of a subset of the bits of the flags parameter describe above. Specifically, it may have the DNS, LLMNR and MDNS bits (the latter in IPv4 and IPv6 flavours) set. Each individual bit is set when the protocol applies to a specific interface and is enabled for it. It is unset otherwise. Specifically, a multicast-capable interface in the "UP" state with an IP address is suitable for LLMNR or MulticastDNS, and any interface that is UP and has an IP address is suitable for DNS. Note the relationship of the bits exposed here with the LLMNR and MulticastDNS properties also exposed on the Link interface. The latter expose what is *configured* to be used on the interface, the former expose what is actually used on the interface, taking into account the abilities of the interface.

DNSSECSupported exposes a boolean field that indicates whether DNSSEC is currently configured and in use on the interface. Note that if DNSSEC is enabled on an interface, it is assumed available until it is detected that the configured server does not actually support it. Thus, this property may initially report that DNSSEC is supported on an interface.

DefaultRoute exposes a boolean field that indicates whether the interface will be used as default route for name queries. See SetLinkDefaultRoute() above.

The other properties reflect the state of the various configuration settings for the link which may be set with the various methods calls such as SetDNS() or SetLLMNR().