Introduction BIND 9.9.3rc1 is the first release candidate of BIND 9.9.3. This document summarizes changes from BIND 9.9.2 to BIND 9.9.3rc1. Please see the CHANGES file in the source code release for a complete list of all changes. Download The latest versions of BIND 9 software can always be found on our web site at http://www.isc.org/downloads/all. There you will find additional information about each release, source code, and pre-compiled versions for Microsoft Windows operating systems. Support Product support information is available on http://www.isc.org/services/support for paid support options. Free support is provided by our user community via a mailing list. Information on all public email lists is available at https://lists.isc.org/mailman/listinfo. Security Fixes Prevents named from aborting with a require assertion failure on servers with DNS64 enabled. These crashes might occur as a result of specific queries that are received. (CVE-2012-5688) [RT #30792 / #30996] Prevents a named assert (crash) when using RPZ to generate A records (but not AAAA records) and DNS64 to generate AAAA records from A records. (CVE-2012-5689) [RT #32141] New Features Adds the command-line tool "dnssec-coverage" that checks to make sure that there is no scheduled lapse in key coverage. Requires python. [RT #28098] Adds support for the EUI48 and EUI64 RR types. [RT #33082] Adds support for the RFC 6742 ILNP record types (NID, LP, L32, and L64). [RT #31836] Feature Changes Changes timing of when slave zones send NOTIFY messages after loading a new copy of the zone. They now send the NOTIFY before writing the zone data to disk. This will result in quicker propagation of updates in multi-level server structures. [RT #27242] Adds a way for a specific version of the XML statistics to be requested. HTTP status 404 is returned if the server does not support the requested version. Servers are still limited to supporting only one version, selected at compile time. [RT #32481] Updates the built-in root hints for D.ROOT-SERVERS.NET whose IPv4 address changed to 199.7.91.13 (as of 3rd January 2013). Note that recursive servers running with an older set of root hints will still operate successfully because there are 12 other root servers whose addresses are correct and who will respond during root priming with the new root nameserver RRset. [RT #32164] Adds RFC 6598 reverse zones to the built-in empty zones list: 64.100.IN-ADDR.ARPA ... 127.100.IN-ADDR.ARPA. [RT #31336] Makes available a new XML schema (version 3.0) for the statistics channel that adds query type statistics at the zone level, flattens the XML tree and uses compressed format to optimize parsing. It also includes new XSL that permits charting via the Google Charts API on browsers that support javascript in XSL. To enable, build BIND with "configure --enable-newstats". [RT #30023] "named -V" can now report a source ID string. (This is will be of most interest to developers and troubleshooters). The source ID for ISC's production versions of BIND is defined in the "srcid" file in the build tree and is normally set to the most recent git hash. [RT #31494] Response Policy Zone performance enhancements. New "response-policy" option "min-ns-dots". "nsip" and "nsdname" now enabled by default with RPZ. [RT #32251] Now includes, in the community contribution section, a dynamically-loadable DLZ module: BDBHPT, contributed by Mark Goldfinch. [RT #32549] Bug Fixes Avoids race condition in data structure initialization with accepting new socket connections. [RT #33084] Fixed memory leak when using ECDSA. [RT #32249] Fixed memory leaks in contrib/query-loc. [RT #32960] Fixed resource leaks and a buffer overrun in contrib/zkt. [RT #32960] Correct initialization errors in libdns when built in libexport mode. [RT #33028] Allow max-cache-size and max-acache-size to accept values greater than 4 gigabytes when built with 64-bit integers. "unlimited" still means 4 gigabytes - 1 and "0" still allows truly unlimited cache sizes. [RT #32358] Removed lock contention issues that slowed zone loading times for 9.9.x compared with 9.8.x. Zone loading times are now faster than they were with 9.8.x. [RT #30399] The zone-statistics option now takes three options: "full", "terse", and "none". "yes" is now a synonym for "full". "no" is now a synonym for "terse", which is how it behaved in previous versions. [RT #29165] The default value for the number of UDP dispatchers is now either the number of CPUs or the number of worker threads, whichever is lower. The previous default was the number of worker threads. [RT #30964] Fixed a crash bug with the loading of incomplete configurations including a slave zone with inline-signing and without a file name. [RT #31946] Corrected dnssec-signzone and dnssec-verify behavior with opt-out delegations and NSEC3. [RT #32072] Fixed rendering issues for some statistics with the XML stats channel. [RT #32587] Prevent a crash-on-shutdown race condition. [RT #32777] Fixed glitch in displaying query data when configured with --enable-newstats and no queries have yet been received. [RT #32620] Fixed bug where expired slave zones could fail to rewrite the zone data file after the master is again available. [RT #31276] Fixed a potential crash when adding and deleting keys with rndc. [RT #32506] Fixed a possible crash with Diffie-Hellman generated TSIG keys. [RT #32649] Now supports NAPTR regular expression validation on all platforms. [RT #32688] Increased maximum allowed key size for some algorithms in ddns-confgen and rndc-confgen. [RT #32753] nsupdate could exit with an assertion when the local and remote address families didn't match. [RT #22897] Fixes some potential memory leaks with gssapi usage. [RT #32405] Fixes a couple of linked-list pointer initialization bugs. [RT #32651] dnssec-keygen and dnssec-setttime disallow setting the delete date to be sooner than the inactive date. [RT #31719] Update HSM PKCS#11 patches to openssl to add support for openssl versions 0.9.8x, 1.0.0j, and 1.0.1c. [RT #29749] ddns-confgen now accepts all the TSIG algorithms that it is documented as supporting when generating keys. [RT #31927] Missing 'managed-keys-directory' is now handled better. Prior to this change, when misconfigured, named could loop and consume 100% CPU. [RT #30625] Now only the programs that use the readline library will link with it (nslookup and nsupdate). [RT #29810] When using 'rndc addzone' of a zone with with 'inline-signing yes;' named will first load the unsigned version and then afterwards successfully create the signed version. (Prior to this fix, the addzone would fail). [RT #31960] dnssec-checkds now emits a clear message when records are not found. This change also fixes a minor reporting problem whereby dnssec-checkds incorrectly reported that no DS records had been found for a KSK, despite having found and listed one. In addition, errors in the man pages (referencing the wrong utility) have been remedied. [RT #31968] dnssec-dsfromkey now no longer puts legal whitespace in DS hashes in order to inter-operate better with some overly-strict registrars. [RT #31951] Addresses portability issues (encountered when testing on HPUX) and corrects "rndc signing -nsec3param" to accept the full range of possible values. [RT #31938] Named should no longer die on shutdown if running with 128 UDP dispatches per interface. [RT #31743] Some DNSSEC-related options (update-check-ksk, dnssec-loadkeys-interval, dnssec-dnskey-kskonly) are now accepted in slave zone definitions in named.conf when inline-signing is being used. [RT #31078] Addresses build problems encountered on NetBSD 6.0 (renames the 'bool' parameter to avoid a namespace clash). [RT #31515] When using the zone reload method of importing changes to named with in-line signing, changes to SOA record parameters (other than the serial number alone) in the un-signed zone will now trigger named to update the signed version of the zone. Prior to this fix, if SOA parameters were updated while the server was offline but without any changes also being made to other records in the zone, then those changes would not be picked up when the server was restarted/reloaded. [RT #29272] named-checkconf now detects missing master lists in also-notify clauses. [RT #30810] Improves locking performance when recursing. (This change implements several different strategies for reducing lock contention, specifically relating to the internal structures that are used when handling upstream queries). [RT #28836] When recursing, named now uses multiple dispatch objects for sending upstream queries; this can improve performance on busy multiprocessor systems by reducing lock contention, particularly when the cache hit rate is low. [RT #28605] Handle cases where a port is reserved and cannot be used as the source for a query. [RT #31778] Correct a case where a negative response could incorrectly be flagged as being DNSSEC authenticated when it was not actually authenticated. [RT #32237] Fix missing includes in testing support library that caused it to fail to build on some platforms. [RT #32012] Return correct error code (FORMERR) when presented with malformed requests containing overly long domain names. [RT #29682] Instead of rejecting and logging a FORMERR, named now accepts duplicate singleton records in a DNS query response. (In some situations, query responses may contain duplicates - and whilst this is not technically correct, BIND has been updated to be more tolerant). [RT #32329] When named allocates an initial per-thread stack size, it first checks the operating system's default value, and if specified, uses that. In the situation where it appears that none is provided, it uses an internal default. This default has been increased from 64K to 1M to accommodate operating systems that require a larger initial stack. [RT #32230] The allow-query-on ACL is now processed correctly in all situations. [RT #29486] The configure script now supports and detects libxml2-2.9.x correctly. [RT #32231] When loading a zone file, named now emits a warning if it encounters a non-blank owner name following $ORIGIN. The reason for this is that when parsing a zone file, the blank owner name indicates that the current name (i.e. the name from the previous record that named loaded) should be used, even though $ORIGIN has changed. Particularly when handling subdomains, this can result in those records being unexpectedly loaded with different labels than intended. [RT #31848] Resolves a problem that when answering queries for nonexistent names via wildcard CNAME records, DNSSEC responses could fail to include the NSEC/NSEC3 records proving the lack of a better answer. [RT #21409] Prevents a named abort (assertion fail) during recovery from an out of memory condition. This crash would be encountered in module general: dst_api.c and logged as REQUIRE((&key->refs)->refs == 0). [RT #32131] A new configure option --with-ecdsa has been added to force building with ECDSA, bypassing the script-based checks that this functionality is available in the build environment. The converse, --without-ecdsa, explicitly disables ECDSA support during the BIND build. Both of these options have been added to assist cross-compilation to environments that do (or don't) support ECDSA, overriding the default build behaviour. [RT #32078] XML statistics generated by Windows builds contained incorrectly formatted "boot-time" and "current-time" values. [RT #32044] dig now prints the timezone as part of the timestamp in the "WHEN" line of the output. [RT #2269] Fixes a race condition in acache.c that could cause named to crash if the acache feature was enabled. [RT #31908] Prevents named from consuming high CPU resources when re-signing if all keys are offline. [RT #31916] Addresses compilation issues when using the GNU build VPATH feature. [RT #31879] Fixes a race condition when DNSSEC validation is canceled (e.g. by server shutdown). [RT #31804] Prevents crashes on startup of named, dig and other utilities from 64-bit builds of BIND in the Solaris 11 environment. Compilers inadvertently created a 64-bit-aligned instruction/32-bit-aligned pointer issue in an area of code that is shared between many of the BIND binaries. Copying the timeval structure from control message data before using it prevents this from happening. [RT #31548] Uses IPV6_USE_MIN_MTU (or equivalent) with TCP in addition to UDP. This change addresses TCP query failures that are due to delays in learning the working PMTU when communicating via tunneled IPv6. [RT #31690] Fixes compilation errors when building with ISC_MEM_TRACKLINES or ISC_MEMPOOL_NAMES disabled and also makes ISC_MEM_DEBUG non-optional. [RT #31559] Prevents named from terminating unexpectedly during on very busy high-end servers that are using the additional section cache ("acache-enable yes;"). [RT #31253] When re-signing a zone, dnssec-signzone now removes RRSIG and NSEC records from nodes that used to be in-zone but are now below a zone cut. This situation is most likely to arise following the delegation of a subdomain where the glue (A and AAAA) records for the nameservers used to be included in the parent zone, but other scenarios are also possible. [RT #31556] Silences unnecessarily noisy OpenSSL logging by suppressing some warning messages and moving others to the "dnssec" logging category. Note that the increased logging was introduced by change 3354 (RT #29932). [RT #31497] Implements a collection of minor changes in response to warnings generated by several source code validation utilities. No instances of problems have been reported, but these code changes improve the future reliability and resilience of BIND9. [RT #31484, RT #31626] dig no longer crashes when using +nssearch with +tcp. [RT #25298] OPT records are no longer removed from signed truncated query responses. Receipt of these responses might cause recursive servers to incorrectly identify the sending servers as unable to support EDNS0. [RT #31439] Message 'sucessfully validated after lower casing signer' is now logged at debug level 1 and has been moved to category "dnssec". (The misspelling is also corrected). RT #31414] "host -C" should no longer crash with a core dump if REFUSED is received. This behaviour was an underlying cause of intermittent and often unreproducible crashes which have been experienced by users of the host command. [RT #31381] A DNSKEY lookup that encounters a CNAME will now no longer return SERVFAIL. This failure mode might have been observed in named's logfiles as a resolver format error "CNAME response for DNSKEY RR". [RT #31262] dig now consistently returns NOERROR in TSIG; prior to this change it would occasionally display '0' instead. [RT #31275] Prevents a named hang (due to a violation of lock ordering that can lead to a deadlock between threads) that may occur in some situations when generating new NSEC / NSEC3 chains. [RT #31224] Slave SOA queries now observe "use-v4-udp-ports" and "use-v6-udp-ports" ranges appropriately. Prior to this change the IPv6 port range was applied to all SOA refresh queries. Most of the time this behaviour would be unnoticed because the IPv6 port range is seldom configured separately and defaults to the IPv4 port range. But if an administrator chose to specify a null IPv6 port range ("use-v6-udp-ports { };") on a slave server, SOA refresh queries would be completely disabled. [RT #24173] named could die if a non-existant master list was referenced in an "also-notify" statement. [RT #31004] In some cases, servers were being marked as not supporting EDNS despite not receiving a successful response [RT #30811] Parsing tests for 32 bit integers will now return a range error on systems that support 64-bit longs. This change may impact administrators who have mistakenly been using serial numbers greater than 2**32 in their zone files (for example, using format YYYYMMDDXXXX) and whose zones loaded, but should have been rejected. The loaded zones would have appeared to be functioning correctly, but in some instances could suffer from operational problems (for example, when enabling IXFR). [RT #30232] Silences spurious "deleted from unreachable cache" messages. [RT #30501] When receiving a query with AD=1 named will now behave in the same way as when DO=1 is set when deciding whether to add NS RRsets to the additional section or not. Prior to this change, when a reply was constructed to a query with DO=1 and if the answer section was signed and valid then named wouldn't add untrusted NS RRsets to the additional section. But if with AD=1 (and DO=0) in the query, then it might have added available but untrusted RRsets to the response, at the same time setting AD=0. [RT #30479] Thank You Thank you to everyone who assisted us in making this release possible. If you would like to contribute to ISC to assist us in continuing to make quality open source software, please visit our donations page at http://www.isc.org/supportisc. (c) 2001-2013 Internet Systems Consortium