RPZ rules cannot be fully relied upon until the summary RPZ database is
updated after an "rndc reload". Wait until the relevant message is
logged after an "rndc reload" to prevent false positives in the
"rpzrecurse" system test caused by the RPZ rules not yet being in effect
by the time ns3 is queried.
Minor changes are:
- Replace the "$RNDCCMD dumpdb" logic with "rndc_dumpdb" from
conf.sh.common (it does the same thing).
- Update a comment to match the grep calls below it (comment said the
rest should be expired, while the grep calls indicate that they
are still in the cache, the comment now explains why).
The dotat() function has been changed to send the TAT
query asynchronously, so there's no lock order loop
because we initialize the data first and then we schedule
the TAT send to happen asynchronously.
This breaks following lock-order loops:
zone->lock (dns_zone_setviewcommit) while holding view->lock
(dns_view_setviewcommit)
keytable->lock (dns_keytable_find) while holding zone->lock
(zone_asyncload)
view->lock (dns_view_findzonecut) while holding keytable->lock
(dns_keytable_forall)
Each dns_rpz_zone_t structure keeps a hash table of the names this RPZ
database contains. Here is what happens when an RPZ is updated:
- a new hash table is prepared for the new version of the RPZ by
iterating over it; each name found is added to the summary RPZ
database,
- every name added to the new hash table is searched for in the old
hash table; if found, it is removed from the old hash table,
- the old hash table is iterated over; all names found in it are
removed from the summary RPZ database (because at that point the old
hash table should only contain names which are not present in the
new version of the RPZ),
- the new hash table replaces the old hash table.
When the new version of the RPZ is iterated over, if a given name is
spelled using a different letter case than in the old version of the
RPZ, the new variant will hash to a different value than the old
variant, which means it will not be removed from the old hash table.
When the old hash table is subsequently iterated over to remove
seemingly deleted names, the old variant of the name will still be
there, causing the name to be deleted from the summary RPZ database
(which effectively causes a given rule to be ignored).
The issue can be triggered not just by altering the case of existing
names in an RPZ, but also by adding sibling names spelled with a
different letter case. This is because RBT code preserves case when
node splitting occurs. The end result is that when the RPZ is iterated
over, a given name may be using a different case than in the zone file
(or XFR contents).
Fix by downcasing all names found in the RPZ database before adding them
to the summary RPZ database.
No issues with the glue cache feature have been reported since its
introduction in BIND 9.12. As the rationale for introducing the
"glue-cache" option was to have a safety switch readily available in
case the glue cache turns out to cause problems, it is time to deprecate
the option. Glue cache will be permanently enabled in a future release,
at which point the "glue-cache" option will be made obsolete.
Attaching and detaching handle pointers will make it easier to
determine where and why reference counting errors have occurred.
A handle needs to be referenced more than once when multiple
asynchronous operations are in flight, so callers must now maintain
multiple handle pointers for each pending operation. For example,
ns_client objects now contain:
- reqhandle: held while waiting for a request callback (query,
notify, update)
- sendhandle: held while waiting for a send callback
- fetchhandle: held while waiting for a recursive fetch to
complete
- updatehandle: held while waiting for an update-forwarding
task to complete
control channel connection objects now contain:
- readhandle: held while waiting for a read callback
- sendhandle: held while waiting for a send callback
- cmdhandle: held while an rndc command is running
httpd connections contain:
- readhandle: held while waiting for a read callback
- sendhandle: held while waiting for a send callback
- rename isc_nmsocket_t->tcphandle to statichandle
- cancelread functions now take handles instead of sockets
- add a 'client' flag in socket objects, currently unused, to
indicate whether it is to be used as a client or server socket
some versions of perl failed to run packet.pl because the 'last'
keyword can't be used outside of a loop block. this commit changes
the packet dumping code to a function so we can use 'return' instead.
the tcp system test uses the 'packet.pl' test tool to send a packet
thousands of times. this took a long time because the tool was waiting
for replies and parsing them; however, for that particular test the
replies aren't relevant.
this commit uses non-blocking sockets and moves the reply parsing
outside the send loop, which speeds the system test up substantially.
The "huge.zone" zone can take longer than 100 seconds to load when
running under a sanitizer. Increase the relevant zone load timeout to
prevent intermittent failures of the "rndc" system test.
The CDS/CDNSKEY record will be published when the DS is in the
rumoured state. However, with the introduction of the rndc '-checkds'
command, the logic in the keymgr was changed to prevent the DS
state to go in RUMOURED unless the specific command was given. Hence,
the CDS was never published before it was seen in the parent.
Initially I thought this was a policy approval rule, however it is
actually a DNSSEC timing rule. Remove the restriction from
'keymgr_policy_approval' and update the 'keymgr_transition_time'
function. When looking to move the DS state to OMNIPRESENT it will
no longer calculate the state from its last change, but from when
the DS was seen in the parent, "DS Publish". If the time was not set,
default to next key event of an hour.
Similarly for moving the DS state to HIDDEN, the time to wait will
be derived from the "DS Delete" time, not from when the DS state
last changed.
The 'rndc_checkds' utility now allows "now" as the time when the DS
has been seen in/seen removed from the parent.
Also it uses "KEYX" as the key argument, rather than key id.
The 'rndc_checkds' will retrieve the key from the "KEYX" string. This
makes the call a bit more readable.
This commit has a lot of updates on comments, mainly to make the
system test more readable.
Also remove some redundant signing policy checks (check_keys,
check_dnssecstatus, check_keytimes).
Finally, move key time checks and expected key time settings above
'rndc_checkds' calls (with the new way of testing next key event
times there is no need to do them after 'rndc_checkds', and moving
them above 'rndc_checkds' makes the flow of testing easier to follow.
Add two more arguments to the dnssec-settime tool. '-P ds' sets the
time that the DS was published in the parent, '-D ds' sets the time
that the DS was removed from the parent (these times are not accurate,
but rely on the user to use them appropriately, and as long as the
time is not before actual publication/withdrawal, it is fine).
These new arguments are needed for the kasp system test. We want to
test when the next key event is once a DS is published, and now
that 'parent-registration-delay' is obsoleted, we need a different
approach to reliable test the timings.
The test works as follows:
1. Client wants to resolve unusual ip6.arpa. name:
test1.test2.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. IN TXT
2. Query is sent to ns7, a qmin enabled resolver.
3. ns7 do the first stage in query minimization for the name and send a new
query to root (ns1):
_.1.0.0.2.ip6.arpa. IN A
4. ns1 delegates ip6.arpa. to ns2.good.:
;; AUTHORITY SECTION:
;ip6.arpa. 20 IN NS ns2.good.
;; ADDITIONAL SECTION:
;ns2.good. 20 IN A 10.53.0.2
5. ns7 do a second round in minimizing the name and send a new query
to ns2.good. (10.53.0.2):
_.8.2.6.0.1.0.0.2.ip6.arpa. IN A
6. ans2 delegates 8.2.6.0.1.0.0.2.ip6.arpa. to ns3.good.:
;; AUTHORITY SECTION:
;8.2.6.0.1.0.0.2.ip6.arpa. 60 IN NS ns3.good.
;; ADDITIONAL SECTION:
;ns3.good. 60 IN A 10.53.0.3
7. ns7 do a third round in minimizing the name and send a new query to
ns3.good.:
_.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. IN A
8. ans3 delegates 1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. to ns4.good.:
;; AUTHORITY SECTION:
;1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. 60 IN NS ns4.good.
;; ADDITIONAL SECTION:
;ns4.good. 60 IN A 10.53.0.4
9. ns7 do fourth round in minimizing the name and send a new query to
ns4.good.:
_.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. IN A
10. ns4.good. doesn't know such name, but answers stating it is authoritative for
the domai:
;; ->>HEADER<<- opcode: QUERY, status: NXDOMAIN, id: 53815
...
;; AUTHORITY SECTION:
1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. 60 IN SOA ns4.good. ...
11. ns7 do another minimization on name:
_.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa
sends to ns4.good. and gets the same SOA response stated in item #10
12. ns7 do another minimization on name:
_.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa
sends to ns4.good. and gets the same SOA response stated in item #10.
13. ns7 do the last query minimization name for the ip6.arpa. QNAME.
After all IPv6 labels are exausted the algorithm falls back to the
original QNAME:
test1.test2.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa
ns7 sends a new query with the original QNAME to ans4.
14. Finally ans4 answers with the expected response:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 40969
;; flags: qr aa; QUESTION: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 8192
;; QUESTION SECTION:
;test1.test2.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. IN TXT
;; ANSWER SECTION:
;test1.test2.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.9.0.9.4.1.1.1.1.8.2.6.0.1.0.0.2.ip6.arpa. 1 IN TXT "long_ip6_name"
The test for assertion failure via large TCP packet needs to be repeated
multiple times (we use 300000). This commit fixes the input file to be
properly hexlified and uses the new packet.pl -r feature to send it
300000 times via TCP.
For some tests, we need to send big data streams (for TCP) or repeated
packets (for UDP), this commits adds `-r` option to packet.pl that sends
the same input <repeats> times using the specified protocol.
In order to lower the amount of memory allocated at startup by named
instances used in the BIND system test suite, set the default value of
"max-cache-size" for these to 2 megabytes. The purpose of this change
is to prevent named instances (or even entire virtual machines) from
getting killed by the operating system on the test host due to excessive
memory use.
Remove all "max-cache-size" statements from named configuration files
used in system tests ("checkconf" notwithstanding) to prevent confusion
as the "-T maxcachesize=..." command line option takes precedence over
configuration files.
An implicit default of "max-cache-size 90%;" may cause memory use issues
on hosts which run numerous named instances in parallel (e.g. GitLab CI
runners) due to the cache RBT hash table now being pre-allocated [1] at
startup. Add a new command line option, "-T maxcachesize=...", to allow
the default value of "max-cache-size" to be overridden at runtime. When
this new option is in effect, it overrides any other "max-cache-size"
setting in the configuration, either implicit or explicit. This
approach was chosen because it is arguably the simplest one to
implement.
The following alternative approaches to solving this problem were
considered and ultimately rejected (after it was decided they were not
worth the extra code complexity):
- adding the same command line option, but making explicit
configuration statements have priority over it,
- adding a build-time option that allows the implicit default of
"max-cache-size 90%;" to be overridden.
[1] see commit e24bc324b4
