Bert Hubert [Fri, 11 Nov 2005 11:57:50 +0000 (11:57 +0000)]
big rename, arg() -> ::arg(), bind() -> ::bind()
remove a read from the configuration from out of the loop - we looked stuff up per packet
add pruning of nsspeeds
Bert Hubert [Fri, 11 Nov 2005 11:41:39 +0000 (11:41 +0000)]
move to 'floats' for nameserver statistics - we don't need 80 bits of precision!
remove some spurious logging that was slowing powerdns down - logging that would never be displayed
parametrise namesever stats and query throttling for possible move to other containers
add support for pruning these stats
Bert Hubert [Fri, 11 Nov 2005 11:17:47 +0000 (11:17 +0000)]
Fix 'same level NS referral', as exemplified by the new same-level-referrel regression test.
This solves where the zone contains 'france.example.com NS somewhere.com' and a query comes in for 'france.example.com'
This problem was new in 2.9.19, caused by the cname/wildcard patches
(plus removal of an unused file)
Bert Hubert [Fri, 28 Oct 2005 23:05:40 +0000 (23:05 +0000)]
document bugfix in previous commit
fix named.conf include file problem reported in http://mailman.powerdns.com/pipermail/pdns-dev/2005-October/000361.html
Bert Hubert [Wed, 26 Oct 2005 22:09:46 +0000 (22:09 +0000)]
update TODO list
some cosmetic changes in pdns_recursor
make dnsreplay_mindex compile again
improve 'trailing garbage' error
teach zone content input/output code about spaces in labels
Bert Hubert [Sun, 23 Oct 2005 16:41:15 +0000 (16:41 +0000)]
Final fix for the AXFR problems, closing Debian bug 330184.
Improve incoming AXFR error reporting
document this for the release notes, plus add hints-file there
Bert Hubert [Sun, 23 Oct 2005 14:53:11 +0000 (14:53 +0000)]
add beginning of next generation zoneparser
add error checking to socket close - there are probably some duplicate closes around, causing havoc
add PTR to records that are propely parsed by the pdns_recursor
teach pdns_recursor to read hints from a zone file, "official" root is still in source
fix some --trace output that would look for .-terminated qnames
document hints-file
Bert Hubert [Thu, 20 Oct 2005 21:20:50 +0000 (21:20 +0000)]
fix some bogus possible closes in resolver.cc (which should be eradicated with a vengeance)
add max-tcp-sessions to documentation
fix some trace outout in syncres
remove some bogus windows code
Bert Hubert [Sun, 16 Oct 2005 19:55:38 +0000 (19:55 +0000)]
add provisional release notes for 2.9.19
implement our '.-policy': we always write out zones with a dot in RecordTextWriter and always expect one
in RecordTextReader. This has upset the pdns_recursor, which has been adapted to deal with it.
janitorial fixes to sdig
fix bug: actual queries for CNAMES got wrong RCODE, depending on compiler memory layout
Bert Hubert [Sun, 16 Oct 2005 15:36:51 +0000 (15:36 +0000)]
make ./configure fail if Boost is not installed, and give out helpful tip
add code to send out root referrals for unauth domains
and documentation --send-root-referral
remove some bogus errors on EINTR (there is a remaining bug out there)
prepare MOADNSParser for record types for which only the name is known, and no maker functions
expand and improve EDNS0 parsing code
templatize optString
fix qtype name reporting in pdns_recorsor (regressed during move to MOADNSParser)
make sure outgoing queries are randomised again (regressed etc)
complete AAAA support in MOADNSParser/generator stuff
Bert Hubert [Sun, 16 Oct 2005 11:47:22 +0000 (11:47 +0000)]
add support for properly truncating records in the pdns_recursor again (was removed during upgrade to MOADNS stuff)
we can now rollback records
add support for honouring EDNS0 queries to pdns_recursor - turns out clients don't use it
EDNS0 parsing code added to MOADNSParser
properly copy recursion desired flag in
Bert Hubert [Mon, 10 Oct 2005 21:30:52 +0000 (21:30 +0000)]
fix compilation of dnsreplay
move pdns_recursor cache to 'optString', which uses placement-new hacks
make pdns_recursor set query-response bit
(reminder: make it ignore questions with query-response!)
improve error reporting in pdns_recursor a bit
some more experiments in dnspbench
Bert Hubert [Mon, 10 Oct 2005 19:42:06 +0000 (19:42 +0000)]
the gist is: move pdns_recursor over to new {zone,packet}-{parser,generator} stuff
this means the pdns_recursor does not use DNSPacket anymore (yay)
Commit adds:
d_qtype storage to all DNSRecordContent classes
infrastructure to extract *only* the record for a packet, assuming it came with a qname
add reportAllTypes() if you want to be able to parse all types,
reportBasicTypes() for the pdns_recursor
reportOtherTypes() if you want the rest too
updates Makefile.am so everything compiles again
fix endian issues in packetwriter/parser
remove virtualness/inheritance from the pdns_recursor storage cache (teeny speedup)
Things might be a bit scary in the recursor now.
Bert Hubert [Sat, 8 Oct 2005 21:58:44 +0000 (21:58 +0000)]
make RFC 3597 'Handling of Unknown DNS Resource Record (RR) Types' support compliant
make toPacket virtual & mandatory for all records
add register of zmakers that convert from zone to native DNSRecords
rebase d_place enum so it agrees with rest of pdns - we will have to move to one ENUM
add documentation to MOADNSParser
make sure 'www.powerdns.com' does not come out as 'www.powerdns.com.' and that root comes out as ''
Bert Hubert [Sat, 8 Oct 2005 16:12:49 +0000 (16:12 +0000)]
fix double-srv regression test error
add label compression to dnswriter, make parser, string parser, sting generator ignore that
make mx, cname, ptr, ns, soa do label compression
fix string parser to not add a trailing . when no zone is configured for expansion
add some comments to DNSPacketWriter
add dnspbench to benchmark and test the new infrastructure away from sdig
new infra appears to generate 100.000 simple packets/s on 1.4GHz
remove a warning in misc.cc
Bert Hubert [Tue, 4 Oct 2005 17:55:01 +0000 (17:55 +0000)]
Fix the bug reported in http://mailman.powerdns.com/pipermail/pdns-users/2005-October/002703.html
regarding duplicate SRV records with differing weights being weeded down to one.
Plus add a regression test for this
Bert Hubert [Sun, 2 Oct 2005 18:49:31 +0000 (18:49 +0000)]
add OPT pseudo-record support
teach DNSPacketWriter about differing sections - answer, auth, additional
add convenience addOpt() function
add setRD() to set recursion desired flag
make sure --trace disables --quiet in pdns_recursor
increase buffer size of Socket to deal with huge fragmented DNSSEC packets
Bert Hubert [Sun, 2 Oct 2005 16:39:31 +0000 (16:39 +0000)]
create more xfrType
create xfrTime for DNSSEC timestamps (with our own y2038 bug!)
hook up xfrTime - our output is now equivalent to that of the latest bind dig I for all queries I can try
Bert Hubert [Sun, 2 Oct 2005 13:28:51 +0000 (13:28 +0000)]
add support for blobs and 8 bit ints to the magical symmetrical transfer machinery
add support for DNSSEC records to MOADNSParser! Tested except for DS, can't find a live DS record.
add support for RP and other odd records to MOADNSParser
except for LOC we are complete now
Bert Hubert [Sun, 2 Oct 2005 01:16:44 +0000 (01:16 +0000)]
further reduction in code, a typical DNS record implementation now only
takes a few lines of code because we benefit from the symmetry between packet writing, reading,
record content reading, record content writing
add 2 new regression tests
add an additional zone for testing (test.com)
rename old test zone file to example.com
add some stuff to named.conf so Bind acts more like powerdns for comparison
improve IPv6 'bound to' output, shift to [::]:port
(Thanks to Niels Bakker, Wichert Akkermans and Gerty de Wolf for suggestions)
split out ipv6 vs ipv4 statistics
make record ordering more stable (mostly cosmetic, but makes regression testing easier)
fix sdig - MOADNSParser API changed
make bindbackend2 case-transparent, so a query for www.UpperCase.nl will not get an answer www.uppercase.nl
make AXFR zone ordering more stable
fix bind2-reload-now (or queueReload in general)
fix d_tcp non-updating, which might also have been causing failing axfrs for some people
fix error report for bind2-reload-now with non-existing domain
stop shuffling axfrs (bah)
Move from u_intxx_t to uintxx_t typedefs, which are preferred:
http://lists.freedesktop.org/pipermail/release-wranglers/2004-August/000926.html
http://www.oreillynet.com/pub/a/network/2003/10/07/michael_barr.html