Commit graph

1493 commits

Author SHA1 Message Date
W.C.A. Wijngaards
c0522043f0 - Fix http2 drop handling to clear the postpone_drop state so that
Some checks failed
ci / build (push) Has been cancelled
other streams on the http2 session are not affected by a drop,
  and can clean up properly if also dropped. Fix http2 send reply
  so that when there is a send failure is does not recurse into
  the mesh functions and also does not drop the connection due to
  the condition of one stream.
2025-12-03 14:41:10 +01:00
W.C.A. Wijngaards
b858801feb - Fix to remove http2 stream mesh state when mesh new request is
Some checks are pending
ci / build (push) Waiting to run
dropping the new request.
2025-12-02 15:31:53 +01:00
Robert Edmonds
fceb4e8585
Mesh reply counters (#1374)
* Statistics counter for number of queries dropped by limit on reply addresses

Request list entries can be associated with multiple pending "reply
addresses". Basically each request list entry keeps its own list of
clients that should receive the response once the recursion is finished.
This requires keeping allocations around for each client, and there is
a global limit on the number of *additional* reply addresses that can
be allocated. (Each new request list entry seems to get its own initial
reply address which is not counted against the limit.)

This commit adds a statistics counter "num_queries_replyaddr_limit" that
counts the number of incoming client queries that have been dropped due
to the restriction on allocating additional reply addresses. This allows
distinguishing these drops from other kinds of drops.

* Statistics counter for number of mesh reply entries

Request list entries can be associated with multiple pending "reply
addresses". Since there is a limit on the number of additional reply
addresses that can be allocated which can cause incoming queries to be
dropped if exceeded, it would be nice to be able to track this number.

This commit basically exports the mesh_area's internal counter
`num_reply_addrs` as "threadX.requestlist.current.replies" /
"total.requestlist.current.replies".
2025-11-13 09:33:05 +01:00
W.C.A. Wijngaards
0a15118aff - Fix that when discard timeout drops packet, they are accounted as
less reply addresses in use in the mesh area.
2025-11-12 11:49:04 +01:00
W.C.A. Wijngaards
c8860a5fb6 - Fix to reply with SERVFAIL when the wait-limit is exceeded. 2025-10-15 11:36:29 +02:00
W.C.A. Wijngaards
735c96aac7 - Fix to drop UDP for discard-timeout, but not stream connections. 2025-10-15 11:04:22 +02:00
Yorgos Thessalonikefs
21f02a0865 - Note clearly that 'wait-limit: 0' disables all wait limits.
- 'wait-limit-cookie: 0' can now disable cookie validated wait
  limits.
2025-10-03 16:44:44 +02:00
W.C.A. Wijngaards
feeebc95f8 - Fix for #1344: Fix that respip and dns64 can be enabled at the
same time, the client info is copied for attach_sub and add_sub
  calls. That makes respip work on dns64 synthesized answers, and
  also makes RPZ work with DNS64. The order for the modules is
  module-config: "respip dns64 validator iterator".
2025-09-30 11:28:15 +02:00
W.C.A. Wijngaards
f1fea8dc46 - Fix #1353: auth-zone can not use empty label for $ORIGIN when
http download.
2025-09-29 14:24:31 +02:00
W.C.A. Wijngaards
6cd595a816 - Fix modstack_call_init to use the original string when it has
changed, to call modstack_config with. And skip the changed name
  in the string correctly. Thanks to Jan Komissar.
2025-09-29 11:31:50 +02:00
Yorgos Thessalonikefs
64645e1401 - Avoid calling mesh_detect_cycle_found() when there is no mesh state
to begin with.
2025-09-24 14:30:24 +02:00
Yorgos Thessalonikefs
e2bf773089 Merge branch 'features/no-ttl-zero-cacherep' 2025-09-19 14:56:04 +02:00
W.C.A. Wijngaards
713b5db537 - Fix to print warning for when so-sndbuf setsockopt is not granted. 2025-09-15 16:11:27 +02:00
Yorgos Thessalonikefs
73e408f1d0 A few changes for TTL processing:
- Cached messages that reach 0 TTL are considered expired. This prevents
  Unbound itself from issuing replies with TTL 0 and possibly causing a
  thundering herd at the last second. Upstream replies of TTL 0 still
  get the usual pass-through but they are not considered for caching
  from Unbound or any of its caching modules.
- 'serve-expired-reply-ttl' is changed and is now capped by the original
  TTL value of the record to try and make some sense when replying
  with expired records.
- TTL decoding was updated to adhere to RFC8767 section 4 where a set
  high-order bit means the value is positive instead of 0.
2025-09-15 10:03:35 +02:00
Yorgos Thessalonikefs
d521135f66 Merge branch 'master' into features/no-ttl-zero-cacherep 2025-09-12 15:24:06 +02:00
W.C.A. Wijngaards
4267de87b5 - Fix #1332: CNAME chains are sometimes not followed when RPZs add a
local CNAME rewrite.
2025-09-09 12:34:11 +02:00
W.C.A. Wijngaards
f2f36a2733 - Fix that the zone acquired timestamp is set after the
zonefile is read.
2025-08-22 14:06:51 +02:00
W.C.A. Wijngaards
78d9bcacb6 - Fix #1319: [FR] zone status for Unbound auth-zones. 2025-08-22 12:40:00 +02:00
Roland van Rijswijk-Deij
44ac818f87
Add extra statistic to track the number of signature validation operations (#1289)
* Add extra statistic to track the number of signature validation operations performed by the validator module

* Move validation operation statistic to mesh as suggested

* Fix NULL pointer dereference in case the mesh is not used (and is `NULL`)

Co-authored-by: Wouter Wijngaards <wcawijngaards@users.noreply.github.com>

* Fix NULL pointer dereference on qstate and qstate->env in unit test situation

---------

Co-authored-by: Wouter Wijngaards <wcawijngaards@users.noreply.github.com>
2025-07-12 16:29:38 +02:00
W.C.A. Wijngaards
cb919d5126 - For #1300: implement sock-queue-timeout for FreeBSD as well. 2025-07-03 15:54:33 +02:00
Jose Luis Duran
41c55ffac1
Fix typos (#1299) 2025-07-02 10:50:49 +02:00
W.C.A. Wijngaards
ce72770f61 - Fix rrset cache create allocation failure case. 2025-06-19 16:27:13 +02:00
W.C.A. Wijngaards
a1d68cdc96 - Fix #1296: DNS over QUIC depends on a very outdated version of
ngtcp2. Fixed so it works with ngtcp2 1.13.0 and OpenSSL 3.5.0.
2025-06-19 14:39:45 +02:00
Yorgos Thessalonikefs
9201c75013 - Fix for consistent use of local zone CNAME alias for configured auth
zones. Now it also applies to downstream configured auth zones.
2025-06-17 15:03:29 +02:00
W.C.A. Wijngaards
a8aa1dbbe1 - Fix conditional expressions with parentheses for bitwise and. 2025-06-11 16:42:43 +02:00
W.C.A. Wijngaards
03772d10fb - Change default for so-sndbuf to 1m, to mitigate a cross-layer
issue where the UDP socket send buffers are exhausted waiting
  for ARP/NDP resolution. Thanks to Reflyable for the report.
2025-05-13 15:04:32 +02:00
Yorgos Thessalonikefs
a35ac5d82e - Fix #1284: NULL pointer deref in az_find_nsec_cover() (latent bug)
by adding a log_assert() to safeguard future development.
2025-05-13 11:00:23 +02:00
Yorgos Thessalonikefs
4e23523d1a
Fix auth nsec3 code (#1280)
- Fix NSEC3 code to not break on broken auth zones that include unsigned
  out of zone (above apex) data. Could lead to hang while trying to
  prove a wildcard answer.
  Reported by Dmitrii Kuvaiskii from Amazon Web Services.

- Tests for NSEC3 auth zones with out of zone data.
2025-05-12 14:26:47 +02:00
Yorgos Thessalonikefs
5dd14e2644 - Sync unbound and unbound-checkconf log output for unknown modules. 2025-05-05 14:47:12 +02:00
W.C.A. Wijngaards
16ee7cf944 - Fix for print of connection type in log-replies for dot and doh. 2025-04-10 09:33:51 +02:00
Willem Toorop
a616437338
DNS Error Reporting (RFC 9567) (#902)
* v1 EDER poc

* remove superfluous edns_list_get_option function

* create an EDER configurable

* Hackathon 114

* Fixes for version -04

* Generated configparser and configlexer are not versioned in master anymore

* Remove NOERROR DNS Error Reporting; not part of final RFC.
* Use assigned IANA EDNS0 Option Code for Report-Channel.

* Fix buffer protection and agent domain validity

* Use DNS Error Reporting instead of the eder nickname

* Update documentation.

* Fix typo.

* Bail out early if ede is not present.

* Forget previous EDNS options from upstream; this is what was
  implicitly happening but not deterministacally.

* Don't report LDNS_EDE_OTHER and bail early if there is no reporting
  agent.

* Only do DNS error reporting when a client asked for something that
  went wrong.

* Add an error reporting agent in the parent that should be ignored.

* review feedback.

* fixup for fast reload

* Add 'num.dns_error_reports' to stats and test for it.

---------

Co-authored-by: TCY16 <tom@nlnetlabs.nl>
Co-authored-by: Yorgos Thessalonikefs <yorgos@nlnetlabs.nl>
2025-04-07 10:25:10 +02:00
W.C.A. Wijngaards
7fb05c01c2 - Fix mesh_copy_client_info to omit null contents from copy. 2025-04-04 08:57:24 +02:00
W.C.A. Wijngaards
c2ca679f5c - Fix #1263: Exempt loopback addresses from wait-limit. 2025-04-03 09:45:36 +02:00
Wouter Wijngaards
218f5cfc92
Fast Reload Option (#1042)
* - fast-reload, add unbound-control fast_reload

* - fast-reload, make a thread to service the unbound-control command.

* - fast-reload, communication sockets for information transfer.

* - fast-reload, fix compile for unbound-dnstap-socket.

* - fast-reload, set nonblocking communication to keep the server thread
  responding to DNS requests.

* - fast-reload, poll routine to test for readiness, timeout fails connection.

* - fast-reload, detect loop in sock_poll_timeout routine.

* - fast-reload, send done and exited notification.

* - fast-reload, defines for constants in ipc.

* - fast-reload, ipc socket recv and send resists partial reads and writes and
  can continue byte by byte. Also it can continue after an interrupt.

* - fast-reload, send exit command to thread when done.

* - fast-reload, output strings for client on string list.

* - fast-reload, add newline to terminal output.

* - fast-reload, send client string to remote client.

* - fast-reload, better debug output.

* - fast-reload, print queue structure, for output to the remote client.

* - fast-reload, move print items to print queue from fast_reload_thread struct.

* - fast-reload, keep list of pending print queue items in daemon struct.

* - fast-reload, comment explains in_list for printq to print remainder.

* - fast-reload, unit test testdata/fast_reload_thread.tdir that tests the
  thread output.

* - fast-reload, fix test link for fast_reload_printq_list_delete function.

* - fast-reload, reread config file from disk.

* - fast-reload, unshare forwards, making the structure locked, with an rwlock.

* - fast-reload, for nonthreaded, the unbound-control commands forward,
  forward_add and forward_delete should be distributed to other processes,
  but when threaded, they should not be distributed to other threads because
  the structure is not thread specific any more.

* - fast-reload, unshared stub hints, making the structure locked, with an rwlock.

* - fast-reload, helpful comments for hints lookup function return value.

* - fast-reload, fix bug in fast reload printout, the strlist appendlist routine,
  and printout time statistics after the reload is done.

* - fast-reload, keep track of reloadtime and deletestime and print them.

* - fast-reload, keep track of constructtime and print it.

* - fast-reload, construct new items.

* - fast-reload, better comment.

* - fast-reload, reload the config and swap trees for forwards and stub hints.

* - fast-reload, in forwards_swap_tree set protection of trees with locks.

* - fast-reload, in hints_swap_tree also swap the node count of the trees.

* - fast-reload, reload ipc to stop and start threads.

* - fast-reload, unused forward declarations removed.

* - fast-reload, unit test that fast reload works with forwards and stubs.

* - fast-reload, fix clang analyzer warnings.

* - fast-reload, small documentation entry in unbound-control -h output.

* - fast-reload, printout memory use by fast reload, in bytes.

* - fast-reload, compile without threads.

* - fast-reload, document fast_reload in man page.

* - fast-reload, print ok when done successfully.

* - fast-reload, option for fast-reload commandline, +v verbosity option,
  with timing and memory use output.

* - fast-reload, option for fast-reload commandline, +p does not pause threads.

* - fast-reload, option for fast-reload commandline, +d drops mesh queries.

* - fast-reload, fix to poll every thread with nopause to make certain that
  resources are not held by the threads and can be deleted.

* - fast-reload, fix to use atomic store for config variables with nopause.

* - fast-reload, reload views.

* - fast-reload, when tag defines are different, it drops the queries.

* - fast-reload, fix tag define check.

* - fast-reload, document that tag change causes drop of queries.

* - fast-reload, fix space in documentation man page.

* - fast-reload, copy respip client information to query state, put views tree
  in module env for lookup.

* - fast-reload, nicer respip view comparison.

* - fast-reload, respip global set is in module env.

* - fast-reload, document that respip_client_info acl info is copied.

* - fast-reload, reload the respip_set.

* - fast-reload, document no pause and pick up of use_response_ip boolean.

* - fast-reload, fix test compile.

* - fast-reload, reload local zones.

* Update locking management for iter_fwd and iter_hints methods. (#1054)

fast reload, move most of the locking management to iter_fwd and
iter_hints methods. The caller still has the ability to handle its
own locking, if desired, for atomic operations on sets of different
structs.

Co-authored-by: Wouter Wijngaards <wcawijngaards@users.noreply.github.com>

* - fast-reload, reload access-control.

* - fast-reload, reload access control interface, such as interface-action.

* - fast-reload, reload tcp-connection-limit.

* - fast-reload, improve comments on acl_list and tcl_list swap tree.

* - fast-reload, fixup references to old tcp connection limits in open tcp
  connections.

* - fast-reload, fixup to clean tcp connection also for different linked order.

* - fast-reload, if no tcp connection limits existed, no need to remove
  references for that.

* - fast-reload, document more options that work and do not work.

* - fast-reload, reload auth_zone and rpz data.

* - fast-reload, fix auth_zones_get_mem.

* - fast-reload, fix compilation of testbound for the new comm_timer_get_mem
  reference in remote control.

* - fast-reload, change use_rpz with reload.

* - fast-reload, list changes in auth zones and stop zonemd callbacks for
  deleted auth zones.

* - fast-reload, note xtree is not swapped, and why it is not swapped.

* - fast-reload, for added auth zones, pick up zone transfer and zonemd tasks.

* - fast-reload, unlock xfr when done with transfer pick up.

* - fast-reload, unlock z when picking up the xfr for it during transfer task
  pick up.

* - fast-reload, pick up task changes for added, deleted and modified auth zones.

* - fast-reload, remove xfr of auth zone deletion without tasks.

* - fast-reload, pick up zone transfer config.

* - fast-reload, the main worker thread picks up the transfer tasks and also
  performs setup of the xfer struct.

* - fast-reload, keep writelock on newzone when auth zone changes.

* - fast-reload, change cachedb_enabled setting.

* - fast-reload, pick up edns-strings config.

* - fast-reload, note that settings are not updated.

* - fast-reload, pick up dnstap config.

* - fast-reload, dnstap options that need to be loaded without +p.

* - fast-reload, fix auth zone reload

* - fast-reload, remove debug for auth zone test.

* - fast-reload, fix auth zone reload with zone transfer.

* - fast-reload, fix auth zone reload lock order.

* - fast-reload, remove debug from fast reload test.

* - fast-reload, remove unused function.

* - fast-reload, fix the worker trust anchor probe timer lock acquisition in
  the probe answer callback routine for trust anchor probes.

* - fast-reload, reload trust anchors.

* - fast-reload, fix trust anchor reload lock on autr global data and test
  for trust anchor reload.

* - fast-reload, adjust cache sizes.

* - fast-reload, reload cache sizes when changed.

* - fast-reload, reload validator env changes.

* - fast-reload, reload mesh changes.

* - fast-reload, check for incompatible changes.

* - fast-reload, improve error text for incompatible change.

* - fast-reload, fix check config option compatibility.

* - fast-reload, improve error text for nopause change.

* - fast-reload, fix spelling of incompatible options.

* - fast-reload, reload target-fetch-policy, outbound-msg-retry, max-sent-count
  and max-query-restarts.

* - fast-reload, check nopause config change for target-fetch-policy.

* - fast-reload, reload do-not-query-address, private-address and capt-exempt.

* - fast-reload, check nopause config change for do-not-query-address,
  private-address and capt-exempt.

* - fast-reload, check fast reload not possible due to interface and
  outgoing-interface changes.

* - fast-reload, reload nat64 settings.

* - fast-reload, reload settings stored in the infra structure.

* - fast-reload, fix modstack lookup and remove outgoing-range check.

* - fast-reload, more explanation for config parse failure.

* - fast-reload, reload worker outside network changes.

* - fast-reload, detect incompatible changes in network settings.

* fast-reload, commit test files.

* - fast-reload, fix warnings for call types in windows compile.

* - fast-reload, fix warnings and comm_point_internal for tcp wouldblock calls.

* - fast-reload, extend lock checks for repeat thread ids.

* - fast-reload, additional test cases, cache change and tag changes.

* - fast-reload, fix documentation for auth_zone_verify_zonemd_with_key.

* - fast-reload, fix copy_cfg type casts and memory leak on config parse failure.

* - fast-reload, fix use of WSAPoll.

* Review comments for the fast reload feature (#1259)

* - fast-reload review, respip set can be null from a view.

* - fast-reload review, typos.

* - fast-reload review, keep clang static analyzer happy.

* - fast-reload review, don't forget to copy tag_actions.

* - fast-reload review, less indentation.

* - fast-reload review, don't leak respip_actions when reloading.

* - fast-reload review, protect NULL pointer dereference in get_mem
  functions.

* - fast-reload review, add fast_reload_most_options.tdir to test most
  options with high verbosity when fast reloading.

* - fast-reload review, don't skip new line on long error printouts.

* - fast-reload review, typo.

* - fast-reload review, use new_z for consistency.

* - fast-reload review, nit for unlock ordering to make eye comparison
  with the lock counterpart easier.

* - fast-reload review, in case of error the sockets are already closed.

* - fast-reload review, identation.

* - fast-reload review, add static keywords.

* - fast-reload review, update unbound-control usage text.

* - fast-reload review, updates to the man page.

* - fast-reload, the fast-reload command is experimental.

* - fast-reload, fix compile of doqclient for fast reload functions.

* Changelog comment for #1042
- Merge #1042: Fast Reload. The unbound-control fast_reload is added.
  It reads changed config in a thread, then only briefly pauses the
  service threads, that keep running. DNS service is only interrupted
  briefly, less than a second.

---------

Co-authored-by: Yorgos Thessalonikefs <yorgos@nlnetlabs.nl>
2025-03-31 15:25:24 +02:00
Yorgos Thessalonikefs
bcf162abd7 - The maximum value of a probe rto was not aligned with the
(configurable) infra-cache-max-rtt value. That could result in
  infra-keep-probing not working if an infra-cache-max-rtt value was chosen
  that was below 12000 ms. This fix still uses a default value of 12000
  ms for the probe but caps it to the infra-cache-max-rtt if that is
  lower.
2025-02-19 12:14:59 +01:00
Yorgos Thessalonikefs
72828ff81c - Consider reconfigurations when calculating the still_useful_timeout
for servers in the infrastructure cache.
2025-02-17 15:21:18 +01:00
Yorgos Thessalonikefs
f822042cd0 - Do not open unencrypted channels next to encrypted ones on the same
port.
2025-01-21 15:26:40 +01:00
Yorgos Thessalonikefs
048c193243 - Use the same interface listening port discovery code for all needed
protocols.
- Port to string only when needed before getaddrinfo().
2025-01-21 10:04:30 +01:00
Yorgos Thessalonikefs
e4483bbbd1
Unique DoT and DoH SSL contexts to allow for different ALPN (#1222) 2025-01-20 15:43:44 +01:00
Yorgos Thessalonikefs
f52b2a6ea2 - Add resolver.arpa and service.arpa to the default locally served
zones.
2025-01-14 17:18:32 +01:00
Yorgos Thessalonikefs
7e4f7ec5be
Merge pull request #1214 from NLnetLabs/bugfix/tls-handshake
Use TCP_NODELAY on TLS sockets to speed up the TLS handshake.
2025-01-10 13:53:46 +01:00
Yorgos Thessalonikefs
7559d26c93 - Use TCP_NODELAY on TLS sockets to speed up the TLS handshake. 2025-01-10 12:11:59 +01:00
Yorgos Thessalonikefs
fff9f62a1e
Serve expired cache update fixes (#1174)
- Fixes a regression bug with serve-expired that appeared in 1.22.0
  and would not allow the iterator to update the cache with
  not-yet-validated entries resulting in increased outgoing traffic.

- Treat serve_expired_norec_ttl as a backoff timer for failed updates of expired records.
- Try to use expired answers instead of SERVFAIL if serve-expired is
  enabled even without serve-expired-client-timeout.
- Add suggestion to refresh the cached norec_ttl and expired_ttl when a
  response cannot update the usable expired entry.
2024-12-31 16:28:12 +01:00
Yorgos Thessalonikefs
61d7250b96
Create the SSL_CTX for QUIC before chroot and privilege drop (#1187)
Fixes #1185 by creating the SSL_CTX for QUIC before chroot and
privilege drop, just like the other SSL_CTX creations.

---------

Co-authored-by: Wouter Wijngaards <wcawijngaards@users.noreply.github.com>
2024-12-03 14:20:33 +01:00
Yorgos Thessalonikefs
b4a9c8bb05 - Safeguard alias loop while looking in the cache for expired answers. 2024-12-03 14:10:17 +01:00
Yorgos Thessalonikefs
1512945c79
Merge pull request #1198 from NLnetLabs/bugfix/log-servfail-serve-expired
Fix log-servfail with serve expired and no useful cache contents
2024-12-03 14:02:03 +01:00
Yorgos Thessalonikefs
c55490c1e6 - Fix #1193: log-servfail fails to log host SERVFAIL responses in
Unbound 1.19.2 on Ubuntu 24.04.1 LTS, by not considering cached
  failures when trying to reply with expired data.
2024-12-02 12:28:11 +01:00
Yorgos Thessalonikefs
f46acec35f - For #1189, homogenize the input buffer size for dname_str(). 2024-12-02 11:53:56 +01:00
Yorgos Thessalonikefs
c1e9d7be7f
Merge pull request #1169 from sakateka/lock-free-az-counters
fix: lock-free counters for auth_zone up/down queries
2024-11-19 17:00:01 +01:00
Sergey Kacheev
2c72a4970b
fix: lock-free counters for auth_zone up/down queries 2024-11-19 18:55:31 +03:00