With the netmgr in use, named may start answering queries before zones
are loaded. This can cause transient failures in system tests after
servers are restarted or reconfigured. This commit adds retry loops
and sleep statements where needed to address this problem.
Also incidentally silenced a clang warning.
If a TCP connection fails while attempting to send a query to a server,
the fetch context will be restarted without marking the target server as
a bad one. If this happens for a server which:
- was already marked with the DNS_FETCHOPT_EDNS512 flag,
- responds to EDNS queries with the UDP payload size set to 512 bytes,
- does not send response packets larger than 512 bytes,
and the response for the query being sent is larger than 512 byes, then
named will pointlessly alternate between sending UDP queries with EDNS
UDP payload size set to 512 bytes (which are responded to with truncated
answers) and TCP connections until the fetch context retry limit is
reached. Prevent such query loops by marking the server as bad for a
given fetch context if the advertised EDNS UDP payload size for that
server gets reduced to 512 bytes and it is impossible to reach it using
TCP.
Performing server setup checks using "+tries=3 +time=5" is redundant as
a single query is arguably good enough for determining whether a given
named instance was set up properly. Only use multiple queries with a
long timeout for resolution checks in the "legacy" system test, in order
to significantly reduce its run time (on a contemporary machine, from
about 1m45s to 0m40s).
In the "legacy" system test, in order to make server setup checks more
consistent with each other, add further checks for either presence or
absence of the EDNS OPT pseudo-RR in the responses returned by the
tested named instances.
Extract repeated dig and grep calls into two helper shell functions,
resolution_succeeds() and resolution_fails(), in order to reduce code
duplication in the "legacy" system test, emphasize the similarity
between all the resolution checks in that test, and make the conditions
for success and failure uniform for all resolution checks in that test.
When testing named instances which are configured to drop outgoing UDP
responses larger than 512 bytes, querying with DO=1 may be used instead
of querying for large TXT records as the effect achieved will be
identical: an unsigned response for a SOA query will be below 512 bytes
in size while a signed response for the same query will be over 512
bytes in size. Doing this makes all resolution checks in the "legacy"
system test more similar. Add checks for the TC flag being set in UDP
responses which are expected to be truncated to further make sure that
tested named instances behave as expected.
Sending TCP queries to test named instances with TCP support disabled
should cause dig output to contain the phrase "connection refused", not
"connection timed out", as such instances never open the relevant
sockets. Make sure that the "legacy" system test fails if the expected
phrase is not found in any of the relevant files containing dig output.
Some system tests assume dig's default setings are in effect. While
these defaults may only be silently overridden (because of specific
options set in /etc/resolv.conf) for BIND releases using liblwres for
parsing /etc/resolv.conf (i.e. BIND 9.11 and older), it is arguably
prudent to make sure that tests relying on specific +timeout and +tries
settings specify these explicitly in their dig invocations, in order to
prevent test failures from being triggered by any potential changes to
current defaults.
- add CHANGES note
- update copyrights and license headers
- add -j to the make commands in .gitlab-ci.yml to take
advantage of parallelization in the gitlab CI process
