There is only a single network manager running on top of the loop
manager (except for tests). Refactor the network manager to be a
singleton (a single instance) and change the unit tests, so that the
shorter read timeouts apply only to a specific handle, not the whole
extra 'connect_nm' network manager instance.
All the applications built on top of the loop manager were required to
create just a single instance of the loop manager. Refactor the loop
manager to not expose this instance to the callers and keep the loop
manager object internal to the isc_loop compilation unit.
This significantly simplifies a number of data structures and calls to
the isc_loop API.
The original `ans.pl` server was based on a copy of the one in
`fetchlimit`, so there are some changes:
- The server now only responds with A replies (which is the only thing
needed).
- The incrementing of the IP address goes beyond the least significant
octet (so, after 192.0.2.255 it will yield 192.0.3.0).
The LSP server (using clangd) was always complaining about:
Suspicious string literal, probably missing a comma
for the two Local IPv6 Unicast Addresses strings that spanned
across multiple lines. Disable clang-format for these two lines.
Add \007.no-apex-covering as an owner name so that the cache does
not get primed with a parent NSEC RRset to test the case where
dns_qp_lookup returns ISC_R_NOTFOUND.
The kasp test cases assume that keymgr operations on the zone under test
have been completed before the test is executed. These are typically
quite fast, but the logs need to be explicitly checked for the messages,
otherwise there's a possibility of race conditions causing the
kasp/rollover tests to become unstable.
Call the wait function in all the kasp/rollover tests where it is
expected (which is generally in each test, unless we're dealing with
unsigned zones).
Many of our test cases only use a single NamedInstance from the
`servers` fixture. Introduce `nsX` helper fixtures to simplify these
tests and reduce boilterplate code further.
Specifically, the test no longer has to either define its own variable
to extract a single server from the list, or use the longer
servers["nsX"] syntax. While this may seem minor, the amount of times it
is repeated across the tests justifies the change. It also promotes
using more explicit server identification, i.e. `nsX`, rather than
generic `server`. This also improves the clarity of the tests and may be
helpful in traceback during debugging as well.
Previously, a lot of the checking was re-implemented and duplicated from
check_rollover_step(). Use that function where possible and only
override the needed checks.
Rather than using multiple slightly modified named.conf files, use a
single template which can be rendered differently based on an input
argument -- in this case, csk_roll.
- Use WatchLog.wait_for_sequence() for the configloading test.
- Omit artifacts check, as it seems quite useless for this test case.
- Join all the tests together. The test case is fairly simple here and
this is the easiest way to ensure the log will be in a predictable
state for all tests. Previously, there was no way to ensure
test_configloading_loading() won't be executed after the other tests,
which would render the check moot. It could also be separated into
its own module, but that seems excessive for a simple test case like
this.
- Use jinja2 template for named.conf and remove setup.sh.
- Remove README and put the relevent comment directly next to the test.
- Remove _sh_ from the test filename to uphold the naming convention.
The test is troublesome, because NamedInstance(identifier) expects that
a directory with such a name exists. While it'd be possible to mock
those directories as well, it'd make the doctest overly long and
complex, which isn't justified, given that it's only testing a couple of
options. Turn it into regular documentation instead.
The buffered reading of finished lines deserves its own class to make
its function clearer, rather than bundling it within the WatchLog class.
Co-Authored-By: Michał Kępień <michal@isc.org>
Extend the WatchLog API with a couple of new matching options.
wait_for_sequence() can be used to check a specific sequence of lines
appears in the log file in the given order.
wait_for_all() ensure that all the provided patterns appear in the log
at least once.
Co-authored-by: Colin Vidal <colin@isc.org>
To allow re-use in upcoming functions, isolate the line matching logic
into a separate function. Use an instance-wide deadline attribute, which
is set by the calling function.
Rather than using two distinct functions for matching either one pattern
(wait_for_line()), or any of multiple patterns (wait_for_lines()), use a
single function that handles both in the same way.
Extend the wait_for_line() API:
1. To allow for usage of one or more FlexPatterns, i.e. either plain
strings to be matched verbatim, or regular expressions. Both can be
used interchangeably to provide the caller to write simple and
readable test code, while allowing for increased complexity to allow
special cases.
2. Always return the regex match, which allows the caller to identify
which line was matched, as well as to extract any additional
information, such as individual regex groups.
To simplify usage of multiple wait_for_*() calls, configure the timeout
value for the WatchLog instance, rather than specifying it for each
call.
This is a preparation/cleanup for implementing multiple wait_for_*()
calls in subsequent commits.
Move the line buffering functionality into _readline() to improve the
readability of code. This also allows reading the file contents from
other functions, since the line buffer is now an attribute of the class.
Fix some broken doctest in watchlog.py (no semantic error, but API
slightly changed and broke some output messags). Also add a test for a
missing failure case.
DNSKEY algorithms RSASHA1 and RSASHA-NSEC3-SHA1 and DS digest type
SHA1 are deprecated. Log when these are present in primary zone
files and when generating new DNSKEYs, DS and CDS records.
Previously, upon receiving a query with TSIG, the server would log
an error and timeout. As there is no way to set up the keyring in the
class anyway (and I believe we don't need it), this commit lets such
queries parse but logs the fact that the query has TSIG.
However, there is a bug [1] in dnspython, which causes `make_response`
and `to_wire` to crash on messages constructed by `from_wire` with
`keyring=False`, so the hack with `message.__class__` is needed to work
around this.
This makes just enough changes for the tsig system test to work with
dnspython >= 2.0.0. On older version the server gives up.
[1] https://github.com/rthalley/dnspython/issues/1205
When compiling with meson, it may be easy to forget to compile system
test dependencies before running the tests. In that case, the test
results would be quite incosistent and unpredictable, with some tests
ending up with ERROR, some with FAILURE and others PASS, without a clear
indication that something is off before running the entire machinery.
Add a check to fail early on if the FEATURETEST binary isn't available,
indicating that system test dependencies were most likely not compiled.
When send_done() is called with a ISC_R_CANCELED status (e.g. because
of a signal from ctrl+c), dig can fail to shutdown because
check_if_done() is not called in the branch. Add a check_if_done()
call.
When reusing a TCP connection (because of the '+keepopen' option),
dig detaches from the query after launching it. This can cause a
crash in dig in rare cases when the "receive" callback is called
earlier than the "send" callback.
The '_cancel_lookup()' function detaches a query only if it's
found in the 'lookup->q' list. Before this commit, with one
additional detach happening before recv_done() -> _cancel_lookup()
is called, it didn't cause problems because an earlier _query_detach()
was unlinking the query from 'lookup->q' (because it was the last
reference), so the additional detach and the skipped detach were
undoing each other.
That is unless the "receive" callback was called earlier than the
"send" callback, in which case the additional detach wasn't destroying
the query (and wasn't unlinking it from 'lookup->q') because the "send"
callback's attachment was still there, and so _cancel_lookup() was
trying to "steal" the "send" callback's attachment and causing an
assertion on 'INSIST(query->sendhandle == NULL);'.
Delete the detachment which caused the described situation.
RRset ordering is now an enum inside struct rdataset attributes. This
was done to keep size to of the structure to its original value before
this MR.
I expect zero performance impact but it should be easier to deal with
attributes in debuggers and language servers.
The ns_client_t struct is reset and zero-ed out on every query,
but some fields (query, message, manager) are preserved.
We observe two things:
- The sendbuf field is going to be overwritten anyway, there's
no need to zero it out.
- The fields are copied out when the struct is zero-ed out, and
then copied back in. For the query field (which is 896 bytes)
this is very inefficient.
This commit makes the reset more efficient avoiding to unnecessary
zero-ing and copy.
The code which checks for both IPv4 and IPv6 mixed usage is inherently
unstable, since the address family is chosen randomly for each
connection.
Closes#5406
It's possible to use pytest.mark.flaky, which achieves the exact same
thing as our custom-defined isctest.mark.flaky -- attempts to rerun the
test on failure, but only is flaky package is available.
The test_kasp_case[secondary.kasp] can sometimes fail on freebsd13. It
appears the test gets stuck on some operation which should be very
quick, but for some reason takes at least a few seconds, causing the
cb_ixfr_is_signed() function to time out.
In one of the cases I investigated, it wasn't a query/response that
caused a timeout, but rather some operation in between. The test
attempts to read from a keyfile/statefile, but I see no reason why that
should block.
In any case, try to increase the timeout for the verification, as that
shouldn't hurt. Also allow the test to be re-run on freebsd13, as it's
likely to be caused by some odd behaviour on that platform -- the issue
doesn't appear anywhere else.
The check "unix socket message counts" sometimes fails with "dnstap
output file smaller than expected". This only happens on freebsd13 and
can't be reproduced easily. There was an attempt to decrease the
required file size in the past, but apparently, the issue can still
occur.
The serve_stale test has some inherent instabilities affecting many
different checks. While the failure rate isn't too high (about four
failures in past three weeks of nightlies), it gets ignored, because the
test has been unstable for a very long time.
This removes a leftover check which should've been removed in a prior
change (see #5244). The softhsm2 failures when attempting to delete the
token should be ignored.
Previously, the one-second sleep was unreliable, as it didn't properly
indicate that the rndc reconfig has been processed. The "test 'rndc
reconfig' with a broken config" check would sometimes fail under TSAN
in CI, because the previous rndc reconfig was still ongoing, and the
subsequent rndc reconfig was ignored.
These tests have been unstable under TSAN in the past, but it appears
that the same failure mode can happen outside of TSAN tests as well.
These tests have produced 12 failures combined in the past three weeks
in nightlies.
The fetchlimit test has failed 8 times in the nightly CI over the past
three weeks. That makes the overall failure rate somewhere around 1 %,
which isn't a lot, but is still annoying when lots of testing is going
on.
Rndc test "test 'rndc reconfig' with a broken config" was failing
intermittently.
Wait for 'running' to be logged rather than just using 'sleep 1' before
calling 'rndc reconfig' a second time to get the expected error message
rather than 'reconfig request ignored: already running'.
There are many system tests where we set `dnssec-validation yes;` only
to also set `trust-anchors { };` which effectively disables the
validation.
This commit replaces this convoluted setup with just
`dnssec-validation no;`.
When the interface-interval parser was changed from uint32 parser to
duration parser, the default value stayed at plain 60 which now means 60
seconds instead of 60 minutes. Fix the default value and the
documentation to match the reality.
Root trust anchors are automatically updated as described in RFC5011.
Add a system test which ensures the root DNSKEYs are always queried by
named during startup.
Because this test uses real internet DNS root servers, it is enabled
only when `CI_ENABLE_LIVE_INTERNET_TESTS` is set.
This test doesn't require artifact checking but when bundled in the same
directory with the shell based tests, the `system:clang:tsan` job was
failing non-deterministically.
The extra messages are typically traceback from assertion failures.
Previously, they'd be printed only after all individual test case
results have been printed. That made it difficult to pair the traceback
to the failing test in some cases, as the node information (aka test
name) might not always be present.
Instead, log any extra messages related to a particular test failure
directly after reporting its result, making the failure details more
readily available and easy to connect with a particular test case.
Make sure the queries and responses are logged at the DEBUG level, which
may provide useful information in case of failing tests.
This doesn't seem to significantly increase the overall artifacts size.
Previously, pytest.log.txt files from all system tests would take around
3 MB, with this change, it's around 8 MB).
In some cases, it's useful to log the sent and received DNS messages.
Add options to enable this on demand. Query is only logged the first
time it's sent, since it doesn't change. If response logging is turned
on, then each response is logged, since it might be different every
time.
When multiline message is logged, indent all but the first line (which
will be preceeded by the LOG_FORMAT). This improves the clarity of logs,
as it's immediately clear which lines are regular log output, and which
ones are multiline debug output.
Adjust the isctest.run.cmd() stdout/stderr logging to this new format.
The messages obtained from test results may contain stuff like detailed
failure/error information, tracebacks etc. In many cases, the message
will be empty, in which case it doesn't need to be logged.
For an example, run test with many test cases, e.g.
verify/test_verify.py, and inspect the tail of the pytest.log.txt before
and after this commit.
Use isctest.log logging facility for consistent and predictable logging
output rather than using print(). Remove writes of stderr, as that
output will be logged in the debug log in case the commands called with
isctest.run.cmd() fails.
We skip those by default as:
a) we don't want to stress the upstream servers in every CI pipeline
b) system tests need to be runnable in a isolated environment by default
Commit 5cd6c173ff changed the contents of
the PACKAGE_DESCRIPTION preprocessor macro from " (<description>)" to
just "<description>" and missed a spot while adjusting all uses of this
macro in the code base. Fix formatting for that malformed log message,
emitted upon named startup.
Previous CPU test relied on either missing default named.conf or the
missing permissions to write into its default directory. In short that
default configuration would be unusable with current user. It would hang
indefinitely at cpu test if the named user could write into directory
specified in default configuration.
Change it instead to explicitly try non-existent configuration file.
It will still fail immediately, but will not rely on running user or
presence of file at default configuration file path.
There is an ongoing debate about the usefulness of the extra artifacts
check. While it might be useful to detect unexpected behaviour in some
tests, it feels extraneous in many cases. This change provides a middle
ground by making the artifact checking optional. This might be
especially useful for writing new tests, since the author gets to decide
whether the check is useful -- and can utilize it, or can skip it for
sake of brevity.
The emptyzones system test ran two consecutive "rndc reload" commands
without waiting for the first one to complete. It used to work because
the commands were serialized, but now an rndc reconfig/reload command is
ignored if another one is already running, so the emptyzones test is
more likely to fail.
Fix this problem by waiting for the log message indicating that all the
zones are loaded before attempting the next reload.
Add a new system test which checks named output when starting,
reconfiguring and reloading the server. It checks that the steps where
configuration is loaded, when named enters exclusive mode, and when the
configuration is applied are all logged, and that they occur in the
correct order. This adds a guard/warning to keep the parsing of the
named.conf outside of the exclusive mode.
The configuration file was parsed when named was in exclusive
(i.e. single-threaded) mode and unable to answer queries. Because
the parsing is a self-contained operation, it is now done before
named enters exclusive mode.
This reduces the amount of time named can't answer queries when
reloading the configuration when the configuration file is large.
Note that exclusive mode is still used for applying the
configuration changes to the server.
Also, simplify the configuration logic by parsing the built-in
configuration only once at server start time.
In some rare cases, the softhsm2 utility reports failure to delete the
token directory, despite the token being found. Subsequent attempts to
delete the token again indicate that the token was deleted.
Ignore this cleanup error, as it doesn't prevent our tests from working
properly. There is also an attempt to delete the token before the test
starts which ensures a clean state before the test is executed, in case
there's actually a leftover token.
For duration measurements, i.e. deadlines and timeouts, it's more
suitable to use monotonic time as it's guaranteed to only go forward,
unlike time.time() which can be affected by local clock settings.
Allow use of exception (and by extension, assert statements) in the
called function in order to extract essential debug information about
the type of failure that was encountered.
In case the called function fails to succeed on the last retry and
raised an exception, log it as error and set it as the assert message to
propagate it through the pytest framework.
Create a test scenario where a signed zone is in multiple views and
then a key may be purged. This is a bug case where the key files are
removed by one view and then the other view starts complaining.
Add a zone using DS records that embed the private algorithm
identifier in the digest field. There are 2 DS record for an
unsupported DNSSEC algorithm one of which that doesn't have a
matching DNSKEY. This zone should validate as insecure as the
validator can establish that both DS records are for unsupported
DNSSEC algorithms.
There are 4 tests:
1) a zone using a known private OID. Validations should succeed
and return AD=1.
2) a zone using an unknown private OID. Validation should succeed
and return AD=0 as the DS to DNSKEY has provably unsupported
algorithm.
3) a zone using a known private OID and an extra DS record. Validation
should succeed as there is DS to DNSKEY with a known algorithm
linkage.
4) a zone using an unknown private OID and an extra DS record.
Validation should fail as only one of the DS records can be matched
to a provable unknown algorithm. The algorithm of the second DS
is indeterminate.
Use the existing RSASHA256 and RSASHA512 implementation to provide
working PRIVATEOID example implementations. We are using the OID
values normally associated with RSASHA256 (1.2.840.113549.1.1.11)
and RSASHA512 (1.2.840.113549.1.1.13).
Add support for proposed DS digest types that encode the private
algorithm identifier at the start of the DS digest as is done for
DNSKEY and RRSIG. This allows a DS record to identify the specific
DNSSEC algorithm, rather than a set of algorithms, when the algorithm
field is set to PRIVATEDNS or PRIVATEOID.
- dns_zone_cdscheck() has been extended to extract the key algorithms
from DNSKEY data when the CDS algorithm is PRIVATEOID or PRIVATEDNS.
- dns_zone_signwithkey() has been extended to support signing with
PRIVATEDNS and PRIVATEOID algorithms. The signing record (type 65534)
added at the zone apex to indicate the current state of automatic zone
signing can now contain an additional two-byte field for the DST
algorithm value, when the DNS secalg value isn't enough information.
dns_resolver_algorithm_supported() has been extended so in addition to
an algorithm number, it can also take a pointer to an RRSIG signature
field in which key information is encoded.
DST algorithm and DNSSEC algorithm values are not necessarily the same
anymore: if the DNSSEC algorithm value is PRIVATEOID or PRIVATEDNS, then
the DST algorithm will be mapped to something else. The conversion is
now done correctly where necessary.
- When the algorithm value for a DNSSEC key is set to PRIVATEOID
or PRIVATEDNS, that's a placeholder value indicating that the
real algorithm identifier is encoded into the key or signature
data. That means the DNSKEY algorithm value and the DST algorithm
value may not be identical, so we must now add environment variables
DEFAULT_ALGORITHM_DST_NUMBER, ALTERNATIVE_ALGORITHM_DST_NUMBER
and DISABLED_ALGORITHM_DST_NUMBER to the test suite, with support
for mapping from DST algorithm value to PRIVATEDNS or PRIVATEOID.
- Some test cases use RRSIGs that have been modified to force
validation to fail. When making those modifications, we now
preserve the first part of the signature, so that PRIVATEDNS and
PRIVATEOID algorithm identifier values will still work. (This
assumes that the identifiers are short and fit into the first
base64 block.)
When going insecure, we publish CDS and CDNSKEY DELETE records. Update
the check_apex function to test this.
Also, skip some tests in the 'check_rollover_step()' function. If
we change the DNSSEC Policy, keys that no longer match the policy will
be retired. When this exactly happens is hard to determine, as it
happens on the reconfigure. So for these tests, we skip the key timing
metadata checks.
Also, the zone becomes unsigned, so don't call 'check_zone_is_signed'
in those cases.
These test cases involve a reconfiguration. The first one is a zone
that changes from dynamic to inline-signing. The others are tests that
key lifetimes are updated correctly after changing them.
apply the existing journal file, if any, to the old version of the
database before diffing it against the new version. then, append
the diff to the end of the journal. this allows easy creation of
a journal file with multiple deltas, by running named-makejournal
successively.
- write the journal to a default location (file1.jnl) if it was not
specified on the commandline.
- exit with a clear error message if file2's SOA serial number is
not changed from file1.
move the "makejournal" tool from bin/tests/system to bin/tools
and rename it to "named-makejournal". add a man page. update
tests to use the new file location.
check that `delv +ns` sends iterative queries over both address
families when -4 and -6 are not used, and suppresses queries
appropriately when they are.
`delv +ns` invokes the same code to perform name resolution as `named`,
but it neglected to set up an IPv6 dispatch object first. Consequently,
it was behaving more like `named -4`.
It now sets up dispatch objects for both address families, and performs
resolver queries to both v4 and v6 addresses, except when one of the
address families has been suppressed by using `delv -4` or `delv -6`.
The "keyopts" field of the dns_zone object was added to support
"auto-dnssec"; at that time the "options" field already had most of
its 32 bits in use by other flags, so it made sense to add a new
field.
Since then, "options" has been widened to 64 bits, and "auto-dnssec"
has been obsoleted and removed. Most of the DNS_ZONEKEY flags are no
longer needed. The one that still seems useful (_FULLSIGN) has been
moved into DNS_ZONEOPT and the rest have been removed, along with
"keyopts" and its setter/getter functions.
Meson is a modern build system that has seen a rise in adoption and some
version of it is available in almost every platform supported.
Compared to automake, meson has the following advantages:
* Meson provides a significant boost to the build and configuration time
by better exploiting parallelism.
* Meson is subjectively considered to be better in readability.
These merits alone justify experimenting with meson as a way of
improving development time and ergonomics. However, there are some
compromises to ensure the transition goes relatively smooth:
* The system tests currently rely on various files within the source
directory. Changing this requirement is a non-trivial task that can't
be currently justified. Currently the last compiled build directory
writes into the source tree which is in turn used by pytest.
* The minimum version supported has been fixed at 0.61. Increasing this
value will require choosing a baseline of distributions that can
package with meson. On the contrary, there will likely be an attempt
to decrease this value to ensure almost universal support for building
BIND 9 with meson.
A dns_view_t has a queryonacl property, which is supposed to hold the
ACL matching the configuration "allow-query-on". However the code
parsing this configuration ACL was missing (or removed by mistake?),
hence this property was always NULL. The ACL was still built but
individually for each zone (which checks if the property exists in the
zone definition, view definition, and finally options definition).
It now create the ACL instance at the view level, enabling zones to
share the same (identical) ACL instead of having their own copies.
A "template" statement can contain the same configuration clauses
as a "zone" statement. A "zone" statement can now reference a
template, and all the clauses in that template will be used as
default values for the zone. For example:
template primary {
type primary;
file "$name.db";
initial-file "primary.db";
};
zone example.com {
template primary;
file "different-name.db"; // overrides the template
};
Special tokens can now be specified in a zone "file" option
in order to generate the filename parametrically. The first
instead of "$name" in the "file" option is replaced with the
zone origin, the first instance of "$type" is replaced with the
zone type (i.e., primary, secondary, etc), and the first instance
of "$view" is replaced with the view name..
This simplifies the creation of zones using initial-file templates.
For example:
$ rndc addzone <zonename> \
{ type primary; file "$name.db"; initial-file "template.db"
When loading a primary zone for the first time, if the zonefile
does not exist but an "initial-file" option has been set, then a
new file will be copied into place from the path specified by
"initial-file".
This can be used to simplify the process of adding new zones. For
instance, a template zonefile could be used by running:
$ rndc addzone example.com \
'{ type primary; file "example.db"; initial-file "template.db"; };'
there were some duplicated syntax checks in named_zone_configure()
that are no longer needed, now that we perform those same checks
using isccfg_check_zoneconf().
there were also some syntax checks that were *only* in
named_zone_configure(), which have now been moved to
isccfg_check_zoneconf(). test cases for them have been
added to the checkconf system test.
the function that checks zone syntax in libisccfg was previously
only called when loading named.conf, not when parsing an an
"rndc addzone" or "rndc modzone" command. this has been corrected.
note that some checks are still skipped: those that check for
duplication of filenames, key directories, etc. to fix this, we'd need
to export the symbol tables that are set up when loading named.conf and
preserve them so they could be reused later.
The "run.sh" script, used by "make test", changes the working
directory to the system test directory before executing pytest.
If the test drops hypothesis artifacts while running, this
can cause spurious test failures due to an apparent mismatch
between the contents of the system test directory and the
temporary pytest directory. This has been addressed by having
"run.sh" call pytest from the parent directory instead.
named-rrchecker now parses the braces which support multi-line input
from the beginning of the input rather than only when reading the
data fields of the record.
The odd-looking "\ " escape is required to italicize <character-string>
without italicizing the final "s". See reStructuredText Markup
Specification, sections "Inline markup recognition rules" and "Escaping
Mechanism". Most importantly:
Escaped whitespace characters are removed from the output document
together with the escaping backslash. This allows for character-level
inline markup.
Deduplicate the code for dynamic updates and increase code clarity by
using an actual dns.update.UpdateMessage rather than an undefined
intermediary format passed around as a list of arguments.
Move the 'csk-roll1' and 'csk-roll2' zones to the rollover test dir and
convert CSK rollover tests to pytest.
The DS swap spans multiple steps. Only the first time we should check
if the "CDS is now published" log is there, and only the first time we
should run 'rndc dnssec -checkds' on the keys. Add a new key to the
step dictionary to disable the DS swap checks.
This made me realize that we need to check for "is not None" in case
the value in the dictionary is False. Update check_rollover_step()
accordingly, and also add a log message which step/zone we are currently
checking.
