resolved: add "static RRs" concept by poettering · Pull Request #41213 · systemd/systemd

poettering · 2026-03-20T10:52:37Z

split out of #40980

keszybz · 2026-03-20T11:12:00Z

Please see the comments in #40980 (review).

…esses

let's ensure the name is actually a valid DNS name.

This only parses a small subset of RR types for now, but we can add more later. Covered are the most important RR types: A, AAAA, PTR.

When we are told to reload our configuration also flush out /etc/hosts explicitly. This is particularly relevant since we suppress too frequent reloads, and hence a synchronous way to force a reload is very useful.

This is an extension of the /etc/hosts concept, but can provide any kind of RRs (well, actually, we only parse A/AAAA/PTR for now, but the concept is open for more). Fixes: systemd#17791

poettering · 2026-03-20T15:10:03Z

I addressed all earlier claude and human reviews now. ptal

github-actions · 2026-03-20T15:16:05Z

Claude review of PR #41213 (`f868062`)

Must fix

IPv4 string dispatch test checks wrong variable — src/test/test-json.c:1708 — The string-based dispatch test dispatches to &dummy but asserts on data.addr.s_addr (which still holds the value from the earlier test). Should check dummy.s_addr instead.
IPv6 string dispatch test checks wrong variable — src/test/test-json.c:1824 — Same issue: dispatches to &dummy but asserts on data.addr.s6_addr instead of dummy.s6_addr.
Missing resolved.conf.xml docs for ReadStaticRecords= — src/resolve/resolved-gperf.gperf:34 — The new config option is registered in gperf and conf.in but has no <varlistentry> in man/resolved.conf.xml. Every other Resolve.* option has a man page entry.

Suggestions

Static records vs /etc/hosts precedence undocumented — src/resolve/resolved-dns-query.c:1069 — Static records are checked before /etc/hosts in query resolution. This priority isn't documented in the systemd.rr(5) man page and could surprise users.
Man page missing /usr/local/lib/ path — man/systemd.rr.xml:31 — The synopsis lists /etc/, /run/, /usr/lib/ paths but omits /usr/local/lib/systemd/resolve/static.d/*.rr. The code uses CONF_PATHS_NULSTR() which includes it, and other man pages (e.g. resolved.conf.xml) consistently list all four paths.

Nits

Missing assert(stats) in load_static_record_file — src/resolve/resolved-static-records.c:82 — The function asserts cf and records but not stats, which is also required non-NULL.
See Also missing resolved.conf(5) — man/systemd.rr.xml:86 — The ReadStaticRecords= option to enable/disable this feature lives in resolved.conf(5), but it's not referenced.
Log message uses raw type number — src/shared/dns-rr.c:2612 — The "Decoding DNS type %i" message could include the symbolic name via dns_type_to_string() for easier debugging.

Workflow run

github-actions · 2026-03-20T15:53:40Z

src/test/test-json.c

+                                },
+                                /* flags= */ 0,
+                                &dummy));
+        ASSERT_EQ(be32toh(data.addr.s_addr), 0xC0A80101U);


Claude: must-fix: This assertion checks data.addr.s_addr, but the string dispatch above writes to &dummy (at offset 0). Since data.addr still holds 0xC0A80101U from the earlier array-based test, this assertion always passes regardless of whether string dispatch works correctly.

Should be:

ASSERT_EQ(be32toh(dummy.s_addr), 0xC0A80101U);

github-actions · 2026-03-20T15:53:41Z

src/test/test-json.c

+                                /* flags= */ 0,
+                                &dummy));
+
+        ASSERT_EQ(data.addr.s6_addr[15], 1);


Claude: must-fix: Same issue as the IPv4 test: the string dispatch writes to &dummy, but these assertions check data.addr.s6_addr (which still holds ::1 from the earlier array-based test). The string dispatch result in dummy is never verified.

Should be:

ASSERT_EQ(dummy.s6_addr[15], 1); for (size_t i = 0; i < 15; i++) ASSERT_EQ(dummy.s6_addr[i], 0);

github-actions · 2026-03-20T15:53:42Z

src/resolve/resolved-gperf.gperf

 Resolve.Cache,                     config_parse_dns_cache_mode,          DNS_CACHE_MODE_YES,  offsetof(Manager, enable_cache)
 Resolve.DNSStubListener,           config_parse_dns_stub_listener_mode,  0,                   offsetof(Manager, dns_stub_listener_mode)
 Resolve.ReadEtcHosts,              config_parse_bool,                    0,                   offsetof(Manager, read_etc_hosts)
+Resolve.ReadStaticRecords,         config_parse_bool,                    0,                   offsetof(Manager, read_static_records)


Claude: must-fix: The new Resolve.ReadStaticRecords= configuration option is registered here and in resolved.conf.in, but there is no corresponding <varlistentry> documentation in man/resolved.conf.xml. Every other Resolve.* option (e.g., ReadEtcHosts=) has a man page entry describing its behavior and default value.

github-actions · 2026-03-20T15:53:43Z

src/resolve/resolved-dns-query.c

            q->state != DNS_TRANSACTION_NULL)
                return 0;

+        r = dns_query_try_static_records(q);


Claude: suggestion: Static records are checked before /etc/hosts in the query resolution order, meaning .rr drop-in files silently take priority over /etc/hosts entries for the same name. This seems intentional, but the systemd.rr(5) man page doesn't mention this precedence relationship. A brief note in the Description section (and/or a comment here analogous to the one above dns_query_try_etc_hosts) would help users understand the resolution order.

github-actions · 2026-03-20T15:53:43Z

man/systemd.rr.xml

+      <member><filename>/run/systemd/resolve/static.d/*.rr</filename></member>
+      <member><filename>/usr/lib/systemd/resolve/static.d/*.rr</filename></member>
+    </simplelist></para>
+  </refsynopsisdiv>


Claude: suggestion: The synopsis lists /etc/, /run/, and /usr/lib/ paths but omits /usr/local/lib/systemd/resolve/static.d/*.rr. The code uses CONF_PATHS_NULSTR() which includes /usr/local/lib/ as well. Other systemd man pages (e.g., resolved.conf.xml) consistently list all four paths.

github-actions · 2026-03-20T15:53:44Z

src/resolve/resolved-static-records.c

+
+        assert(cf);
+        assert(records);
+


Claude: nit: Missing assert(stats) -- the function asserts cf and records but not the stats double-pointer parameter, which is dereferenced unconditionally on the next line.

github-actions · 2026-03-20T15:53:45Z

man/systemd.rr.xml

+    <title>See Also</title>
+    <para><simplelist type="inline">
+      <member><citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry></member>
+      <member><citerefentry><refentrytitle>systemd-resolved.service</refentrytitle><manvolnum>8</manvolnum></citerefentry></member>


Claude: nit: The See Also section references systemd-resolved.service(8) and hosts(5) but does not reference resolved.conf(5), which is where the ReadStaticRecords= option to enable/disable this feature lives.

github-actions · 2026-03-20T15:53:46Z

src/shared/dns-rr.c

+                *ret = TAKE_PTR(rr);
+        return 0;
+}
+


Claude: nit: The log message "Decoding DNS type %i is currently not supported." outputs just the raw numeric type. Consider including the symbolic name via dns_type_to_string() for easier debugging, similar to the pattern in resolvectl.c which uses both the name and number.

poettering added resolve please-review PR is ready for (re-)review by a maintainer labels Mar 20, 2026

github-actions bot added documentation build-system util-lib tests meson labels Mar 20, 2026

keszybz added reviewed/needs-rework 🔨 PR has been reviewed and needs another round of reworks and removed please-review PR is ready for (re-)review by a maintainer labels Mar 20, 2026

poettering added 6 commits March 20, 2026 16:01

json-util: optionally accept string-based serialization for IPv4 addr…

ce0a59a

…esses

json-util: add json_dispatch_in6_addr()

f19f889

dns-rr: tighten rules on parsing RR keys from JSON

9ba3a6d

let's ensure the name is actually a valid DNS name.

dns-rr: add dns_resource_record_from_json()

212e69a

This only parses a small subset of RR types for now, but we can add more later. Covered are the most important RR types: A, AAAA, PTR.

resolved: also flush /etc/hosts on reload

5b5edcd

When we are told to reload our configuration also flush out /etc/hosts explicitly. This is particularly relevant since we suppress too frequent reloads, and hence a synchronous way to force a reload is very useful.

resolved: add ability to define additional local RRs via drop-ins

f868062

This is an extension of the /etc/hosts concept, but can provide any kind of RRs (well, actually, we only parse A/AAAA/PTR for now, but the concept is open for more). Fixes: systemd#17791

poettering force-pushed the resolved-static-rrs branch from c80680e to f868062 Compare March 20, 2026 15:08

poettering added please-review PR is ready for (re-)review by a maintainer claude-review and removed reviewed/needs-rework 🔨 PR has been reviewed and needs another round of reworks labels Mar 20, 2026

github-actions bot reviewed Mar 20, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

resolved: add "static RRs" concept#41213

resolved: add "static RRs" concept#41213
poettering wants to merge 6 commits intosystemd:mainfrom
poettering:resolved-static-rrs

poettering commented Mar 20, 2026

Uh oh!

keszybz commented Mar 20, 2026

Uh oh!

poettering commented Mar 20, 2026

Uh oh!

github-actions bot commented Mar 20, 2026 •

edited

Loading

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

github-actions bot Mar 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

poettering commented Mar 20, 2026

Uh oh!

keszybz commented Mar 20, 2026

Uh oh!

poettering commented Mar 20, 2026

Uh oh!

github-actions bot commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Claude review of PR #41213 (f868062)

Must fix

Suggestions

Nits

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

github-actions bot commented Mar 20, 2026 •

edited

Loading

Claude review of PR #41213 (`f868062`)