The x509.subject_base() function is only used in tests. During
the recent certificate refactoring, we had to get rid of the
ipalib.x509 import from the module scope so that there were no
circular dependecies and add it exactly to this funcion which
is not used in the production code.
Reviewed-By: Tibor Dudlak <tdudlak@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
We don't need the strip_header() function, to load an unknown
x509 certificate, load_unknown_x509_certificate() should be used.
Reviewed-By: Tibor Dudlak <tdudlak@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
As man page already does it, update the help text to show REPLICA_FILE
as optional.
Fixes https://pagure.io/freeipa/issue/7223
Signed-off-by: Rishabh Dave <rishabhddave@gmail.com>
Reviewed-By: Abhijeet Kasurde <akasurde@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
ipa-replica-conncheck can be called with --principal / --password or
with an existing Kerberos credential cache in order to supply the
authorized identity logging in to the master machine (in
auto-master-check mode).
In domain-level 0, the tool is called with --principal and password
and tries to obtain a TGT by performing kinit, but does not set the
env var KRB5CCNAME. Subsequent calls to IPA API do not use the
credential cache and fail. In this case, ipa-replica-conncheck falls
back to using SSH to check master connectivity instead of IPA API,
and the ssh check is less robust.
The code should set the KRB5CCNAME env var for IPA API to use the
credential cache.
Fixes:
https://pagure.io/freeipa/issue/7221
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Fixed ipa-restore code to get rid of bytes related TypeError and
to get ipa-restore work again.
https://pagure.io/freeipa/issue/7131
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
krb5-1.16 includes DAL version 7, which changes the signature of
audit_as_req to include local and remote address parameters.
This patch just enables building against the new DAL version and bumps
the minimum in freeipa.spec.in, but doesn't use the new information
for anything.
Reviewed-By: Tomas Krizek <tkrizek@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
LDIFWriter.unparse() expects the scalar values of the attributes
of the entries to be bytes as it applies a byte regular expression
to check whether to base64-encode the values or not. Previously,
we were passing the scalar attribute values as strings which
was breaking the LDIFWriter.unparse() exectution.
https://pagure.io/freeipa/issue/7131
Reviewed-By: Felipe Barreto <fbarreto@redhat.com>
The new template should allow to use python3 to run ipa-run-tests
since it provides the required dependencies for HTML test results
extraction and python3-paramiko.
Reviewed-By: Felipe Barreto <fbarreto@redhat.com>
The content synchronization plugin can be limited to the dns subtree in
Directory Server. This increases performance and helps to prevent some
potential issues.
Fixes: https://pagure.io/freeipa/issue/6515
Signed-off-by: Tomas Krizek <tkrizek@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Before applying replica-s3u2proxy.ldif, we check
if the values are already there. The values can be
there if a replica installation was done in the past
and some info was left behind. Also, the code checks
the values independently.
https://pagure.io/freeipa/issue/7174
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
cffi throws warnings during the implicit cast from char * to
unsigned char * since the support of these casts is nearing
its end of life.
https://pagure.io/freeipa/issue/7131
Previously, CSRs were handled as a Str parameter which brought
trouble to Python 3 because of its more strict type requirements.
We introduce a CertificateSigningRequest parameter which allows to
use python-cryptography x509.CertificateSigningRequest to represent
CSRs in the framework.
https://pagure.io/freeipa/issue/7131
The type checks in ipalib.parameters were too strict. An object
that inherits from a type should implement its public interface.
This should allow us checking for types of objects whose class
implementations are private to a module but they implement a certain
public interface (which is typical for e.g. python-cryptography).
https://pagure.io/freeipa/issue/7131
cert_get_requestdata() method is meant for internal use only and
is never passed a file. Make its parameter public_key_info Bytes
to better represent what's actually being passed to it.
https://pagure.io/freeipa/issue/7131
To avoid insidious bug during server installation on Fedora 27,
the dependency of 389-ds-base is bumped.
https://bugzilla.redhat.com/show_bug.cgi?id=1488295
Signed-off-by: Tomas Krizek <tkrizek@redhat.com>
Reviewed-By: Thierry Bordaz <tbordaz@redhat.com>
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
During integration tests, the log level of 8192 (replication debugging)
was excessive and made reading 389-ds logs very hard without providing
any useful information.
Part of: https://pagure.io/freeipa/issue/7162
Signed-off-by: Tomas Krizek <tkrizek@redhat.com>
Reviewed-By: Thierry Bordaz <tbordaz@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Fixing import errors introduced by commits
icac3475a0454b730d6e5b2093c2e63d395acd387 and
0b7d9c5.
https://pagure.io/freeipa/issue/7132
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Tomas Krizek <tkrizek@redhat.com>
Replace custom file_exists() and dir_exists() functions with proper
functions from Python's stdlib.
The change also gets rid of pylint's invalid bad-python3-import error,
https://github.com/PyCQA/pylint/issues/1565
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
cmd.api.Command.dnsrecord_split_parts expects name to be unicode
string and instead gets ascii. It leads to an error:
ipa: ERROR: invalid 'name': must be Unicode text
This commit's change is casting name's type to unicode so
'ipa dnsrecord-mod' will not fail with error above.
https://pagure.io/freeipa/issue/7185
Reviewed-By: Tibor Dudlak <tdudlak@redhat.com>
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
When performing force synchronization, implicitly ignore initial
time skew (if any) and restore it afterwards.
This also changes semantics of force-sync by waiting until the end of
the initial replication.
Fixes https://pagure.io/freeipa/issue/7211
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Initial replica creation can go with ignoring time skew checks.
We should, however, force time skew checks during normal operation.
Fixes https://pagure.io/freeipa/issue/7211
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
When running on memory-constrained systems, the `ipa-server-install`
program often fails during the "Configuring certificate server
(pki-tomcatd)" stage in FreeIPA 4.5 and 4.6.
The memory-intensive dogtag service causes swapping on low-memory
systems right after start-up, and especially new certificate
operations requested via certmonger can exceed the dbus client default
25 second timeout.
This patch changes dbus client timeouts for some such operations to
120 seconds (from the default 25 seconds, IIRC).
See more discussion in FreeIPA PR #1078 [1] and FreeIPA container
issue #157 [2]. Upstream ticket at [3].
[1]: https://github.com/freeipa/freeipa/pull/1078
[2]: https://github.com/freeipa/freeipa-container/issues/157
[3]: https://pagure.io/freeipa/issue/7213
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
If the function `install_kra` or `install_ca` fails
on call `host.run_command(command, raiseonerr=raiseonerr)`
then the logs are not collected.
This situation is not optimal because we need to see what happend
during the debbuging the tests.
So, this patch solves this situation and it adds try--finally
construction.
https://pagure.io/freeipa/issue/7214
Reviewed-By: Tibor Dudlak <tdudlak@redhat.com>
The scenario switching from externally signed CA to self-signed CA is
currently failing because the certmonger helper goes through the wrong
code path when the cert is not self-signed.
When the cert is not self-signed but the admin wants to switch to self-signed
a new cert needs to be requested, not retrieved from LDAP.
https://pagure.io/freeipa/issue/7173
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
This fix adds correct usage of host.hostname in logger.
Fixes: https://pagure.io/freeipa/issue/7190
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Some dependencies like Dogtag's pki.client library and custodia use
python-requsts to make HTTPS connection. python-requests prefers
PyOpenSSL over Python's stdlib ssl module. PyOpenSSL is build on top
of python-cryptography which trigger a execmem SELinux violation
in the context of Apache HTTPD (httpd_execmem).
When requests is imported, it always tries to import pyopenssl glue
code from urllib3's contrib directory. The import of PyOpenSSL is
enough to trigger the SELinux denial.
Block any import of PyOpenSSL's SSL module in wsgi by raising an
ImportError. The block is compatible with new python-requests with
unbundled urllib3, too.
Fixes: https://pagure.io/freeipa/issue/5442
Fixes: RHBZ#1491508
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Tomas Krizek <tkrizek@redhat.com>
This causes Firefox to report our CA certificate as not-trustworthy.
We were previously doing this correctly, however it slipped as an
error due to certificate refactoring.
https://pagure.io/freeipa/issue/7210
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
User would like to see CA installation command in KRA installation
warning message.
This makes warning message similar to other installer messages where it
does suggests a command to run.
https://pagure.io/freeipa/issue/6952
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
Reviewed-By: Felipe Barreto <fbarreto@redhat.com>
Previously IPA would set the LimitNOFILE value to 8192 to increase
the number of concurrent clients. 389-ds-base does this by default
as of 1.3.7.0.
Remove the IPA-specific tuning and rely on the out-of-the-box
389-ds-base tuning.
Bump the required version of 389-ds-base to 1.3.7.0.
Any other tuning added by 389-ds-base will result in a
dirsrv.systemd.rpmsave file which admins will need to merge
in manually, like typical .rpmsave config changes.
https://pagure.io/freeipa/issue/6994
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
Now, the criteria option is working for both commands
and the commands are able to handle with wrong input values.
https://pagure.io/freeipa/issue/7134
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
If LDAP or HTTP Server Cert are not issued by ipa ca, they are not tracked.
In this case, it is not necessary to add them to the tracking requests list.
https://pagure.io/freeipa/issue/7151
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
ipa-server-upgrade needs to configure certmonger with the right options
in order to track PKI, HTTP and LDAP certs (for instance the RA agent cert
location has changed from older releases).
The upgrade code looks for existing tracking requests with the expected
options by using criteria (location of the NSSDB, nickname, CA helper...)
If a tracking request is not found, it means that it is either using wrong
options or not configured. In this case, the upgrade stop tracking
all the certs, reconfigures the helpers, starts tracking the certs so that
the config is up-to-date.
The issue is that the criteria is using the keyword 'ca' instead of
'ca-name' and this leads to upgrade believing that the config needs to be
updated in all the cases.
https://pagure.io/freeipa/issue/7151
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Tests should always rely on api.env.* values when possible.
Without this running the tests remotely can result in errors such
as ldap{search,modify,passwd} attempting to connect to the
wrong URI and failing.
https://fedorahosted.org/freeipa/ticket/6622
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
The message about an error during replica setup was causing the
users to think the installation gone wrong even though this was
an expected behavior when ipa-replica-install was ran without
--no-pkinit flag and CA somehow is not reachable which defines
that there is something wrong in a topology but does not lead
to failure of the replica's installation. So now installation
will not print error messages to stdout but rather will give a
recomendation to user and write the old error message to log
as a warning so it still will be easy to find if needed.
https://pagure.io/freeipa/issue/7179
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
While processing enterprise principals the information about trusted domains
might not be up-to-date. With this patch ipadb_reinit_mspac() is called if an
unknown domain is part of the enterprise principal.
Resolves https://pagure.io/freeipa/issue/7172
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
This patch is mainly for test_forced_client_reenrolment suite
where when we are not in control of our client DNS we create an
overlap zone in order to get the host records updated. This also
sets resolv.conf before every ipa-client-install to the ipa master.
https://pagure.io/freeipa/issue/7124
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Milan Kubik <mkubik@redhat.com>
When running tests in python3 we get bytes object instead of
bytestring from get_file_contents() and when passing it to
run_command() we later fail on concatenation in shell_quote().
https://pagure.io/freeipa/issue/7131
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
This is causing python2 tests print ugly warnings about the
deprecation of the `message` attribute in python2.6.
https://pagure.io/freeipa/issue/7131
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
If the # of group memberships exceeded the search size limit
then SizeLimitExceeded was raised. Being in too many groups
should not cause a *_show to fail.
https://pagure.io/freeipa/issue/7112
Reviewed-By: Stanislav Laznicka <slaznick@redhat.com>
