Home | History | Annotate | Line # | Download | only in changelog
      1 .. Copyright (C) Internet Systems Consortium, Inc. ("ISC")
      2 ..
      3 .. SPDX-License-Identifier: MPL-2.0
      4 ..
      5 .. This Source Code Form is subject to the terms of the Mozilla Public
      6 .. License, v. 2.0.  If a copy of the MPL was not distributed with this
      7 .. file, you can obtain one at https://mozilla.org/MPL/2.0/.
      8 ..
      9 .. See the COPYRIGHT file distributed with this work for additional
     10 .. information regarding copyright ownership.
     11 
     12 BIND 9.20.24
     13 ------------
     14 
     15 Security Fixes
     16 ~~~~~~~~~~~~~~
     17 
     18 - Fix DNS64 owner case after DNAME restart. ``4de2229364``
     19 
     20   When BIND 9 is configured to use DNS64 and encounters a DNAME
     21   redirect, it could end up using freed memory for the DNS response
     22   owner name. This caused the response to contain corrupted data. This
     23   fix ensures the correct owner name is used when constructing the
     24   synthesized response after a DNAME redirect.
     25 
     26   ISC thanks Qifan Zhang of Palo Alto Networks for reporting the issue.
     27   :gl:`#5934`
     28 
     29 New Features
     30 ~~~~~~~~~~~~
     31 
     32 - Enable PR-Agent reviews on merge requests. ``46e4c236a3``
     33 
     34   Adds a CI job that runs PR-Agent against each merge request opened
     35   from the canonical repository, posting an automated review and
     36   code-improvement suggestions as MR comments. The job is gated to
     37   same-project source branches so the OpenAI key and personal access
     38   token are not exposed to fork pipelines. :gl:`!12034`
     39 
     40 Removed Features
     41 ~~~~~~~~~~~~~~~~
     42 
     43 - Remove ineffective TCP fallback after repeated UDP timeouts.
     44   ``eb13adcb47``
     45 
     46   When an authoritative server failed to respond to two consecutive UDP
     47   queries, named marked the next retry as TCP but still sent it over
     48   UDP, producing misleading dnstap records. The ineffective retry path
     49   has been removed; a corrected TCP fallback will be restored in future
     50   BIND 9 versions. :gl:`#5529` :gl:`!12049`
     51 
     52 - Remove useless PR-Agent jobs. ``8851b279d0``
     53 
     54   The experiment was a failure, the PR-Agent doesn't send a full context
     55   to the AI Agents and the results are abysmal because of that.
     56   :gl:`!12120`
     57 
     58 Feature Changes
     59 ~~~~~~~~~~~~~~~
     60 
     61 - Fall back to TCP on a UDP response with a mismatched query id.
     62   ``b2367aaea2``
     63 
     64   BIND used to wait silently for the correct DNS message id on a UDP
     65   fetch even after receiving a response from the expected server with
     66   the wrong id, leaving room for off-path spoofing attempts to keep
     67   guessing within that window.  The resolver now retries the fetch over
     68   TCP on the first such response, and a new MismatchTCP statistics
     69   counter tracks how often the fallback fires. :gl:`#5449` :gl:`!12025`
     70 
     71 - Limit the number of glue records cached from a referral.
     72   ``eb401f6b92``
     73 
     74   When a delegation response contained many glue addresses per listed
     75   nameserver, all of them were cached without a per-nameserver bound,
     76   inflating resolver cache memory beyond what resolution could ever use.
     77   The cache now keeps at most 20 IPv4 and 20 IPv6 glue addresses per
     78   nameserver from a delegation. :gl:`#5701` :gl:`!11972`
     79 
     80 - Fix a resolver stall on a CNAME response to a DS query. ``1407f48670``
     81 
     82   A validating resolver could stall for about twelve seconds and then
     83   return SERVFAIL when an authoritative server answered a DS query with
     84   a CNAME. Such responses are now rejected promptly, so the query fails
     85   fast instead of hanging. :gl:`#5878` :gl:`!12147`
     86 
     87 - Named could crash on concurrent TKEY DELETE for the same key.
     88   ``dd02abde67``
     89 
     90   On a server configured with tkey-gssapi-keytab (or
     91   tkey-gssapi-credential), an authenticated peer could crash named by
     92   sending two TKEY DELETE requests for the same dynamic key in rapid
     93   succession.  This has been fixed. :gl:`#6001` :gl:`!12042`
     94 
     95 Bug Fixes
     96 ~~~~~~~~~
     97 
     98 - The resolver now removes other RRsets at the same name when caching a
     99   CNAME. ``1547447491``
    100 
    101   When an RRset is in stale cache, and the authoritative server changes
    102   the record type to CNAME, the resolver fails to refresh the stale
    103   cache. This has been fixed. :gl:`#5302` :gl:`!12040`
    104 
    105 - Fix nxdomain-redirect combined with dns64. ``95274e9455``
    106 
    107   When a resolver was configured with both `nxdomain-redirect` and
    108   `dns64` in the same view, an AAAA query for a nonexistent name could
    109   abort `named`. The combination failed whenever the redirect zone held
    110   A records but no AAAA records.  The server now serves the empty AAAA
    111   response from the redirect zone as-is, instead of attempting DNS64
    112   synthesis on top of it. :gl:`#5789` :gl:`!12123`
    113 
    114 - Clear REDIRECT flag when it isn't needed. ``86bb27060c``
    115 
    116   When `nxdomain-redirect` is in use, and a recursive query is used to
    117   get the redirected answer, a flag is set to distinguish it from a
    118   normal recursive response. Previously, that flag was left set
    119   afterward, which could trigger an assertion if a normal recursive
    120   query was sent later on behalf of the same client: for example,
    121   because the `filter-aaaa` plugin was in use.  This has been fixed.
    122   :gl:`#5936` :gl:`!12076`
    123 
    124 - Fix data race during rndc dumpdb or zone load. ``947e5c7983``
    125 
    126   'rndc dumpdb' against a server with zones, and async zone load, had a
    127   timing window where the operation's completion could fire before the
    128   server had finished registering the operation, occasionally leading to
    129   a possible crash.  The completion is now delivered after the
    130   registration is in place. :gl:`#5952` :gl:`!12021`
    131 
    132 - Bound memory use during incoming zone transfers. ``5d7f241fdf``
    133 
    134   During an incoming zone transfer, an optimization could let the batch
    135   of pending records grow without bound for a large zone, raising memory
    136   usage. It gave no measurable performance benefit, so it has been
    137   removed. :gl:`#5958` :gl:`!12142`
    138 
    139 - Disable output escaping in bind9.xsl. ``b514e663eb``
    140 
    141   The statistics charts where not displaying on some browsers. This has
    142   been fixed. :gl:`#5990` :gl:`!12019`
    143 
    144 - Fix crash on badly configured secondary signer. ``edc1ef084f``
    145 
    146   A badly configured secondary signer that was missing the 'file' entry
    147   caused the server to crash, rather than to reject the configuration.
    148   This has been fixed. :gl:`#5993` :gl:`!12112`
    149 
    150 - Avoid named assertion failure during parent-NS lookups when none
    151   exist. ``5c0c4786dd``
    152 
    153   Configuring the root zone as a signed primary with parental agents (or
    154   with notify-on-cds-changes) caused named to exit on an internal
    155   assertion as soon as the DS-publication machinery tried to look up the
    156   parent NS RRset  the root has no parent. The lookup is now
    157   short-circuited cleanly.
    158 
    159   Similar, a zone with no NS records in the parent caused named to exit
    160   in the same way. :gl:`#5910` :gl:`#5996` :gl:`!12053`
    161 
    162 - Reject RRSIG records covering meta-types. ``7517e39504``
    163 
    164   A recursive resolver could accept and cache an RRSIG record whose
    165   Type-Covered field names a meta-type (ANY, AXFR, IXFR, MAILA, MAILB),
    166   even though no real RRset of those types ever exists. Such records are
    167   now rejected by the DNS message parser. :gl:`#6002` :gl:`!12051`
    168 
    169 - Validate nsec3hash arguments instead of relying on atoi()
    170   ``a59080c053``
    171 
    172   The nsec3hash tool parsed its algorithm, flags, and iterations
    173   arguments with atoi(), then range-checked the result. For values that
    174   overflow int during digit-by-digit accumulation, atoi() is undefined;
    175   in practice on musl libc the modular wrap leaves n == 0, which
    176   silently passes the "iterations > 0xffffU" check. On Alpine Linux this
    177   made nsec3hash succeed with iterations treated as 0 for inputs like
    178   4294967296 (2^32).
    179 
    180   The latent bug only surfaced when the recent image rebuild pulled in
    181   Hypothesis 6.152.9 (2026-05-19), which unified the distribution used
    182   for bounded and unbounded integers() strategies. The new smoother
    183   distribution explores the 2^32 boundary on unbounded ranges like
    184   integers(min_value=65536); earlier versions did not reach there, so
    185   test_nsec3hash_too_many_iterations only started failing on Alpine
    186   after the image refresh.
    187 
    188   Replace the three atoi() calls with isc_parse_uint8 /
    189   isc_parse_uint16, which uniformly reject overflow, trailing garbage,
    190   leading sign, and non-numeric input across libc implementations. As a
    191   side effect, error messages now include the offending argument and a
    192   specific reason ("out of range" vs "not a valid number").
    193 
    194   Assisted-by: Claude:claude-opus-4-7 :gl:`#6013` :gl:`!12074`
    195 
    196 - Fix wrong variable in named_server_sync() log message. ``09f9889ef6``
    197 
    198   named_server_sync() logged isc_result_totext(result) but returns
    199   tresult. The loop accumulates errors into tresult, so result only
    200   holds the last iteration's value. If the last view succeeded but an
    201   earlier one failed, the log would incorrectly say "success".
    202   :gl:`!12156`
    203 
    204 - Refine resolver fetch loop detection. ``787faa02a7``
    205 
    206   The resolver's fetch loop detection now triggers only when a new fetch
    207   would join an already in-flight fetch that is also one of its own
    208   ancestors, which is the actual loop condition.  Previously the check
    209   ran against the original request before the fetch was set up.
    210   :gl:`!12146`
    211 
    212 
    213