1SYSTEMD-RESOLVED.SERVICE(8)systemd-resolved.serviceSYSTEMD-RESOLVED.SERVICE(8)
2
6 systemd-resolved.service, systemd-resolved - Network Name Resolution
7 manager
8
10 systemd-resolved.service
11 /usr/lib/systemd/systemd-resolved
13
15 systemd-resolved is a system service that provides network name
16 resolution to local applications. It implements a caching and
17 validating DNS/DNSSEC stub resolver, as well as an LLMNR and
18 MulticastDNS resolver and responder. Local applications may submit
19 network name resolution requests via three interfaces:
20 • The native, fully-featured API systemd-resolved exposes on the bus,
22 see org.freedesktop.resolve1(5) and org.freedesktop.LogControl1(5)
23 for details. Usage of this API is generally recommended to clients
24 as it is asynchronous and fully featured (for example, properly
25 returns DNSSEC validation status and interface scope for addresses
26 as necessary for supporting link-local networking).
27 • The glibc getaddrinfo(3) API as defined by RFC3493[1] and its
29 related resolver functions, including gethostbyname(3). This API is
30 widely supported, including beyond the Linux platform. In its
31 current form it does not expose DNSSEC validation status
32 information however, and is synchronous only. This API is backed by
33 the glibc Name Service Switch (nss(5)). Usage of the glibc NSS
34 module nss-resolve(8) is required in order to allow glibc's NSS
35 resolver functions to resolve hostnames via systemd-resolved.
36 • Additionally, systemd-resolved provides a local DNS stub listener
38 on the IP addresses 127.0.0.53 and 127.0.0.54 on the local loopback
39 interface. Programs issuing DNS requests directly, bypassing any
40 local API may be directed to this stub, in order to connect them to
41 systemd-resolved. Note however that it is strongly recommended that
42 local programs use the glibc NSS or bus APIs instead (as described
43 above), as various network resolution concepts (such as link-local
44 addressing, or LLMNR Unicode domains) cannot be mapped to the
45 unicast DNS protocol.
46 The DNS stub resolver on 127.0.0.53 provides the full feature set
48 of the local resolver, which includes offering LLMNR/MulticastDNS
49 resolution. The DNS stub resolver on 127.0.0.54 provides a more
50 limited resolver, that operates in "proxy" mode only, i.e. it will
51 pass most DNS messages relatively unmodified to the current
52 upstream DNS servers and back, but not try to process the messages
53 locally, and hence does not validate DNSSEC, or offer up
54 LLMNR/MulticastDNS. (It will translate to DNS-over-TLS
55 communication if needed however.)
56 The DNS servers contacted are determined from the global settings in
58 /etc/systemd/resolved.conf, the per-link static settings in
59 /etc/systemd/network/*.network files (in case systemd-
60 networkd.service(8) is used), the per-link dynamic settings received
61 over DHCP, information provided via resolvectl(1), and any DNS server
62 information made available by other system services. See
63 resolved.conf(5) and systemd.network(5) for details about systemd's own
64 configuration files for DNS servers. To improve compatibility,
65 /etc/resolv.conf is read in order to discover configured system DNS
66 servers, but only if it is not a symlink to
67 /run/systemd/resolve/stub-resolv.conf, /usr/lib/systemd/resolv.conf or
68 /run/systemd/resolve/resolv.conf (see below).
69
71 systemd-resolved synthesizes DNS resource records (RRs) for the
72 following cases:
73 • The local, configured hostname is resolved to all locally
75 configured IP addresses ordered by their scope, or — if none are
76 configured — the IPv4 address 127.0.0.2 (which is on the local
77 loopback interface) and the IPv6 address ::1 (which is the local
78 host).
79 • The hostnames "localhost" and "localhost.localdomain" as well as
81 any hostname ending in ".localhost" or ".localhost.localdomain" are
82 resolved to the IP addresses 127.0.0.1 and ::1.
83 • The hostname "_gateway" is resolved to all current default routing
85 gateway addresses, ordered by their metric. This assigns a stable
86 hostname to the current gateway, useful for referencing it
87 independently of the current network configuration state.
88 • The hostname "_outbound" is resolved to the local IPv4 and IPv6
90 addresses that are most likely used for communication with other
91 hosts. This is determined by requesting a routing decision to the
92 configured default gateways from the kernel and then using the
93 local IP addresses selected by this decision. This hostname is only
94 available if there is at least one local default gateway
95 configured. This assigns a stable hostname to the local outbound IP
96 addresses, useful for referencing them independently of the current
97 network configuration state.
98 • The hostname "_localdnsstub" is resolved to the IP address
100 127.0.0.53, i.e. the address the local DNS stub (see above) is
101 listening on.
102 • The hostname "_localdnsproxy" is resolved to the IP address
104 127.0.0.54, i.e. the address the local DNS proxy (see above) is
105 listening on.
106 • The mappings defined in /etc/hosts are resolved to their configured
108 addresses and back, but they will not affect lookups for
109 non-address types (like MX). Support for /etc/hosts may be disabled
110 with ReadEtcHosts=no, see resolved.conf(5).
111
113 The lookup requests that systemd-resolved.service receives are routed
114 to the available DNS servers, LLMNR, and MulticastDNS interfaces
115 according to the following rules:
116 • Names for which synthetic records are generated (the local
118 hostname, "localhost" and "localdomain", local gateway, as listed
119 in the previous section) and addresses configured in /etc/hosts are
120 never routed to the network and a reply is sent immediately.
121 • Single-label names are resolved using LLMNR on all local interfaces
123 where LLMNR is enabled. Lookups for IPv4 addresses are only sent
124 via LLMNR on IPv4, and lookups for IPv6 addresses are only sent via
125 LLMNR on IPv6. Note that lookups for single-label synthesized names
126 are not routed to LLMNR, MulticastDNS or unicast DNS.
127 • Queries for the address records (A and AAAA) of single-label
129 non-synthesized names are resolved via unicast DNS using search
130 domains. For any interface which defines search domains, such
131 look-ups are routed to the servers defined for that interface,
132 suffixed with each of those search domains. When global search
133 domains are defined, such look-ups are routed to the global
134 servers. For each search domain, queries are performed by suffixing
135 the name with each of the search domains in turn. Additionally,
136 lookup of single-label names via unicast DNS may be enabled with
137 the ResolveUnicastSingleLabel=yes setting. The details of which
138 servers are queried and how the final reply is chosen are described
139 below. Note that this means that address queries for single-label
140 names are never sent out to remote DNS servers by default, and
141 resolution is only possible if search domains are defined.
142 • Multi-label names with the domain suffix ".local" are resolved
144 using MulticastDNS on all local interfaces where MulticastDNS is
145 enabled. As with LLMNR, IPv4 address lookups are sent via IPv4 and
146 IPv6 address lookups are sent via IPv6.
147 • Queries for multi-label names are routed via unicast DNS on local
149 interfaces that have a DNS server configured, plus the globally
150 configured DNS servers if there are any. Which interfaces are used
151 is determined by the routing logic based on search and route-only
152 domains, described below. Note that by default, lookups for domains
153 with the ".local" suffix are not routed to DNS servers, unless the
154 domain is specified explicitly as routing or search domain for the
155 DNS server and interface. This means that on networks where the
156 ".local" domain is defined in a site-specific DNS server, explicit
157 search or routing domains need to be configured to make lookups
158 work within this DNS domain. Note that these days, it's generally
159 recommended to avoid defining ".local" in a DNS server, as
160 RFC6762[2] reserves this domain for exclusive MulticastDNS use.
161 • Address lookups (reverse lookups) are routed similarly to
163 multi-label names, with the exception that addresses from the
164 link-local address range are never routed to unicast DNS and are
165 only resolved using LLMNR and MulticastDNS (when enabled).
166 If lookups are routed to multiple interfaces, the first successful
168 response is returned (thus effectively merging the lookup zones on all
169 matching interfaces). If the lookup failed on all interfaces, the last
170 failing response is returned.
171 Routing of lookups is determined by the per-interface routing domains
173 (search and route-only) and global search domains. See
174 systemd.network(5) and resolvectl(1) for a description how those
175 settings are set dynamically and the discussion of Domains= in
176 resolved.conf(5) for a description of globally configured DNS settings.
177 The following query routing logic applies for unicast DNS lookups
179 initiated by systemd-resolved.service:
180 • If a name to look up matches (that is: is equal to or has as
182 suffix) any of the configured routing domains (search or
183 route-only) of any link, or the globally configured DNS settings,
184 "best matching" routing domain is determined: the matching one with
185 the most labels. The query is then sent to all DNS servers of any
186 links or the globally configured DNS servers associated with this
187 "best matching" routing domain. (Note that more than one link might
188 have this same "best matching" routing domain configured, in which
189 case the query is sent to all of them in parallel).
190 In case of single-label names, when search domains are defined, the
192 same logic applies, except that the name is first suffixed by each
193 of the search domains in turn. Note that this search logic doesn't
194 apply to any names with at least one dot. Also see the discussion
195 about compatibility with the traditional glibc resolver below.
196 • If a query does not match any configured routing domain (either
198 per-link or global), it is sent to all DNS servers that are
199 configured on links with the DefaultRoute= option set, as well as
200 the globally configured DNS server.
201 • If there is no link configured as DefaultRoute= and no global DNS
203 server configured, one of the compiled-in fallback DNS servers is
204 used.
205 • Otherwise the unicast DNS query fails, as no suitable DNS servers
207 can be determined.
208 The DefaultRoute= option is a boolean setting configurable with
210 resolvectl or in .network files. If not set, it is implicitly
211 determined based on the configured DNS domains for a link: if there's a
212 route-only domain other than "~.", it defaults to false, otherwise to
213 true.
214 Effectively this means: in order to support single-label
216 non-synthesized names, define appropriate search domains. In order to
217 preferably route all DNS queries not explicitly matched by routing
218 domain configuration to a specific link, configure a "~." route-only
219 domain on it. This will ensure that other links will not be considered
220 for these queries (unless they too carry such a routing domain). In
221 order to route all such DNS queries to a specific link only if no other
222 link is preferred, set the DefaultRoute= option for the link to true
223 and do not configure a "~." route-only domain on it. Finally, in order
224 to ensure that a specific link never receives any DNS traffic not
225 matching any of its configured routing domains, set the DefaultRoute=
226 option for it to false.
227 See org.freedesktop.resolve1(5) for information about the D-Bus APIs
229 systemd-resolved provides.
230
232 This section provides a short summary of differences in the resolver
233 implemented by nss-resolve(8) together with systemd-resolved and the
234 traditional stub resolver implemented in nss-dns.
235 • Some names are always resolved internally (see Synthetic Records
237 above). Traditionally they would be resolved by nss-files if
238 provided in /etc/hosts. But note that the details of how a query is
239 constructed are under the control of the client library. nss-dns
240 will first try to resolve names using search domains and even if
241 those queries are routed to systemd-resolved, it will send them out
242 over the network using the usual rules for multi-label name routing
243 [3].
244 • Single-label names are not resolved for A and AAAA records using
246 unicast DNS (unless overridden with ResolveUnicastSingleLabel=, see
247 resolved.conf(5)). This is similar to the no-tld-query option being
248 set in resolv.conf(5).
249 • Search domains are not used for suffixing of multi-label names.
251 (Search domains are nevertheless used for lookup routing, for names
252 that were originally specified as single-label or multi-label.) Any
253 name with at least one dot is always interpreted as a FQDN.
254 nss-dns would resolve names both as relative (using search domains)
255 and absolute FQDN names. Some names would be resolved as relative
256 first, and after that query has failed, as absolute, while other
257 names would be resolved in opposite order. The ndots option in
258 /etc/resolv.conf was used to control how many dots the name needs
259 to have to be resolved as relative first. This stub resolver does
260 not implement this at all: multi-label names are only resolved as
261 FQDNs.[4]
262 • This resolver has a notion of the special ".local" domain used for
264 MulticastDNS, and will not route queries with that suffix to
265 unicast DNS servers unless explicitly configured, see above. Also,
266 reverse lookups for link-local addresses are not sent to unicast
267 DNS servers.
268 • This resolver reads and caches /etc/hosts internally. (In other
270 words, nss-resolve replaces nss-files in addition to nss-dns).
271 Entries in /etc/hosts have highest priority.
272 • This resolver also implements LLMNR and MulticastDNS in addition to
274 the classic unicast DNS protocol, and will resolve single-label
275 names using LLMNR (when enabled) and names ending in ".local" using
276 MulticastDNS (when enabled).
277 • Environment variables $LOCALDOMAIN and $RES_OPTIONS described in
279 resolv.conf(5) are not supported currently.
280 • The nss-dns resolver maintains little state between subsequent DNS
282 queries, and for each query always talks to the first listed DNS
283 server from /etc/resolv.conf first, and on failure continues with
284 the next until reaching the end of the list which is when the query
285 fails. The resolver in systemd-resolved.service however maintains
286 state, and will continuously talk to the same server for all
287 queries on a particular lookup scope until some form of error is
288 seen at which point it switches to the next, and then continuously
289 stays with it for all queries on the scope until the next failure,
290 and so on, eventually returning to the first configured server.
291 This is done to optimize lookup times, in particular given that the
292 resolver typically must first probe server feature sets when
293 talking to a server, which is time consuming. This different
294 behaviour implies that listed DNS servers per lookup scope must be
295 equivalent in the zones they serve, so that sending a query to one
296 of them will yield the same results as sending it to another
297 configured DNS server.
298
300 Four modes of handling /etc/resolv.conf (see resolv.conf(5)) are
301 supported:
302 • systemd-resolved maintains the
304 /run/systemd/resolve/stub-resolv.conf file for compatibility with
305 traditional Linux programs. This file lists the 127.0.0.53 DNS stub
306 (see above) as the only DNS server. It also contains a list of
307 search domains that are in use by systemd-resolved. The list of
308 search domains is always kept up-to-date. Note that
309 /run/systemd/resolve/stub-resolv.conf should not be used directly
310 by applications, but only through a symlink from /etc/resolv.conf.
311 This file may be symlinked from /etc/resolv.conf in order to
312 connect all local clients that bypass local DNS APIs to
313 systemd-resolved with correct search domains settings. This mode of
314 operation is recommended.
315 • A static file /usr/lib/systemd/resolv.conf is provided that lists
317 the 127.0.0.53 DNS stub (see above) as only DNS server. This file
318 may be symlinked from /etc/resolv.conf in order to connect all
319 local clients that bypass local DNS APIs to systemd-resolved. This
320 file does not contain any search domains.
321 • systemd-resolved maintains the /run/systemd/resolve/resolv.conf
323 file for compatibility with traditional Linux programs. This file
324 may be symlinked from /etc/resolv.conf and is always kept
325 up-to-date, containing information about all known DNS servers.
326 Note the file format's limitations: it does not know a concept of
327 per-interface DNS servers and hence only contains system-wide DNS
328 server definitions. Note that /run/systemd/resolve/resolv.conf
329 should not be used directly by applications, but only through a
330 symlink from /etc/resolv.conf. If this mode of operation is used
331 local clients that bypass any local DNS API will also bypass
332 systemd-resolved and will talk directly to the known DNS servers.
333 • Alternatively, /etc/resolv.conf may be managed by other packages,
335 in which case systemd-resolved will read it for DNS configuration
336 data. In this mode of operation systemd-resolved is consumer rather
337 than provider of this configuration file.
338 Note that the selected mode of operation for this file is detected
340 fully automatically, depending on whether /etc/resolv.conf is a symlink
341 to /run/systemd/resolve/resolv.conf or lists 127.0.0.53 as DNS server.
342
344 SIGUSR1
345 Upon reception of the SIGUSR1 process signal systemd-resolved will
346 dump the contents of all DNS resource record caches it maintains,
347 as well as all feature level information it learnt about configured
348 DNS servers into the system logs.
349 SIGUSR2
351 Upon reception of the SIGUSR2 process signal systemd-resolved will
352 flush all caches it maintains. Note that it should normally not be
353 necessary to request this explicitly – except for debugging
354 purposes – as systemd-resolved flushes the caches automatically
355 anyway any time the host's network configuration changes. Sending
356 this signal to systemd-resolved is equivalent to the resolvectl
357 flush-caches command, however the latter is recommended since it
358 operates in a synchronous way.
359 SIGRTMIN+1
361 Upon reception of the SIGRTMIN+1 process signal systemd-resolved
362 will forget everything it learnt about the configured DNS servers.
363 Specifically any information about server feature support is
364 flushed out, and the server feature probing logic is restarted on
365 the next request, starting with the most fully featured level. Note
366 that it should normally not be necessary to request this explicitly
367 – except for debugging purposes – as systemd-resolved automatically
368 forgets learnt information any time the DNS server configuration
369 changes. Sending this signal to systemd-resolved is equivalent to
370 the resolvectl reset-server-features command, however the latter is
371 recommended since it operates in a synchronous way.
372
374 systemd-resolved supports the service credentials logic as implemented
375 by ImportCredential=/LoadCredential=/SetCredential= (see
376 systemd.exec(1) for details). The following credentials are used when
377 passed in:
378 network.dns, network.search_domains
380 May contain a space separated list of DNS server IP addresses and
381 DNS search domains. This information is only used when no explicit
382 configuration via /etc/systemd/resolved.conf, /etc/resolv.conf or
383 the kernel command line has been provided.
384
386 systemd-resolved also honours two kernel command line options:
387 nameserver=, domain=
389 Takes the IP address of a DNS server (in case of nameserver=), and
390 a DNS search domain (in case of domain=). May be used multiple
391 times, to define multiple DNS servers/search domains. If either of
392 these options are specified /etc/resolv.conf will not be read and
393 the DNS= and Domains= settings of resolved.conf(5) will be ignored.
394 These two kernel command line options hence override system
395 configuration.
396
398 The systemd-resolved service listens on the following IP ports:
399 • Port 53 on IPv4 addresses 127.0.0.53 and 127.0.0.54 (both are on
401 the local loopback interface "lo"). This is the local DNS stub, as
402 discussed above. Both UDP and TCP are covered.
403 • Port 5353 on all local addresses, both IPv4 and IPv6 (0.0.0.0 and
405 ::0), for MulticastDNS on UDP. Note that even though the socket is
406 bound to all local interfaces via the selected "wildcard" IP
407 addresses, the incoming datagrams are filtered by the network
408 interface they are coming in on, and separate MulticastDNS
409 link-local scopes are maintained for each, taking into
410 consideration whether MulticastDNS is enabled for the interface or
411 not.
412 • Port 5355 on all local addresses, both IPv4 and IP6 (0.0.0.0 and
414 ::0), for LLMNR, on both TCP and UDP. As with MulticastDNS
415 filtering by incoming network interface is applied.
416
418 systemd(1), resolved.conf(5), dnssec-trust-anchors.d(5), nss-
419 resolve(8), resolvectl(1), resolv.conf(5), hosts(5),
420 systemd.network(5), systemd-networkd.service(8)
421
423 1. RFC3493
424 https://tools.ietf.org/html/rfc3493
425 2. RFC6762
427 https://tools.ietf.org/html/rfc6762
428 3. For example, if /etc/resolv.conf has
430 nameserver 127.0.0.53
432 search foobar.com barbar.com
433 and we look up "localhost", nss-dns will send the following queries
436 to systemd-resolved listening on 127.0.0.53:53: first
437 "localhost.foobar.com", then "localhost.barbar.com", and finally
438 "localhost". If (hopefully) the first two queries fail, systemd-
439 resolved will synthesize an answer for the third query.
440 When using nss-dns with any search domains, it is thus crucial to
442 always configure nss-files with higher priority and provide
443 mappings for names that should not be resolved using search
444 domains.
445 4. There are currently more than 1500 top-level domain names defined,
447 and new ones are added regularly, often using "attractive" names
448 that are also likely to be used locally. Not looking up multi-label
449 names in this fashion avoids fragility in both directions: a valid
450 global name could be obscured by a local name, and resolution of a
451 relative local name could suddenly break when a new top-level
452 domain is created, or when a new subdomain of a top-level domain in
453 registered. Resolving any given name as either relative or absolute
454 avoids this ambiguity.
455systemd 254 SYSTEMD-RESOLVED.SERVICE(8)