Security Architect role was supposed to manage sysaccount objects. But
since description attribute is missing from the list of the managed
permissions, the role is unable to modify 'description' field. Same for
nsAccountLock which needs an explicit ACI.
Fixes: https://pagure.io/freeipa/issue/9875
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Commands sysaccount-enable/sysaccount-disable allow to lock/unlock the
account. This is exposed via operational attribute `nsAccountLock`
which has to be explicitly requested to query the state. However, if
account wasn't explicitly disabled before, `nsAccountLock` will not be
returned by the LDAP server.
Ensure the nsaccountlock attribute is always retrieved, validated, and
normalized across sysaccount operations.
- Invoke validate_nsaccountlock in create and modify pre-callbacks to
enforce valid values
- Invoke convert_nsaccountlock in create, modify, list, reset, and
policy post-callbacks to normalize the nsaccountlock attribute and
default to False in case it is absent.
Fixes: https://pagure.io/freeipa/issue/9842
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Thomas Woerner <twoerner@redhat.com>
Introduce support for LDAP-based system accounts by adding a dedicated
sysaccount plugin with full CLI commands, extending role membership to
include system accounts, and providing warnings for passsync
configuration updates across servers, along with corresponding
documentation and tests.
New Features:
Implement system accounts LDAP object with CLI commands for add,
delete, modify, find, show, enable, and disable.
Enhancements:
Extend role and baseldap plugins to support sysaccount membership
and manage passsync managers with a ServerSysacctMgrUpdateRequired
warning.
Documentation:
Add a design document for system accounts and regenerate API docs
for sysaccount commands and role membership.
Tests:
Update existing role and service plugin tests to cover sysaccount
membership.
Fixes: https://pagure.io/freeipa/issue/9842
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rafael Guterres Jeffman <rjeffman@redhat.com>
Reviewed-By: Thomas Woerner <twoerner@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
Instead of modifying /etc/krb5.conf with all IPA realm configuration, follow
the practice of creating a separate config snippet file owned by IPA.
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
In order to generate the private key for a a LWCA (subca)
on an HSM the name of the subca needs to be the HSM token
name : name_of_subca, e.g. ipa_token:test.
This works fine now without any code changes but it requires
that admins always remember to include the prefix which is
unlikely (everyone makes mistakes). So do it for them if
an HSM is present and a token is not provided. We only support
one token at a time in IPA so for now this is sufficient.
One can also mix-and-match including the token and not.
For example you can run:
$ ipa ca-add test --subject cn=test
$ ipa ca-show ipa_token:test
It shouldn't be an issue if a lwca name contains a colon.
Fixes: https://pagure.io/freeipa/issue/9865
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Rafael Guterres Jeffman <rjeffman@redhat.com>
This repalces the straight string parsing from IPAv3.0.0.
It should be much more robust and result in a valid
configuration file when complete. aug.save() will
assure that as it will raise an error otherwise.
This will handle the following cases:
* db_library = ipadb.so : do nothing
* db_library is missing, set to ipadb.so
* db_library is something else, replace wih ipadb.so
Fixes: https://pagure.io/freeipa/issue/9862
Related: https://pagure.io/freeipa/issue/5913
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
We already have libpwquality plumbed in. The ask here was
to support the quality credits as well.
There are four credits: digits, uppers, lowers and others
These either count complexity towards satisfying the minimum
length or set the minimum number of a type allowed.
If any of the four credits is configured then it will cause
the password policy plugin to go through the libpwquality.
Reminder that libpwquality sets its own min password length
floor that should be overridden already.
If min_classes is set as well then that is still enforced.
Fixes: https://pagure.io/freeipa/issue/9835
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: David Hanina <dhanina@redhat.com>
Adds a small information note, regarding launching modern webui with
redirect link.
Signed-off-by: David Hanina <dhanina@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
If a subCA only exists in LDAP then it couldn't be removed.
Allow the process to continue. It will fail in the same way if
it doesn't exist on both Dogtag and IPA.
Related: https://pagure.io/freeipa/issue/9738
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
LDAP expects case-insensitive DNs, so modify LDIF parser to
compare DNs as case-insensitive strings and use case-preserving but
case-insensitive dictionary.
Fixes: https://pagure.io/freeipa/issue/9854
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Samba 4.23 ignores trust to PIM and 'within the forest' domains. It
didn't know what to do with them in past as well but starting with 4.23
the domain trust entries which have trust attributes with
LSA_TRUST_ATTRIBUTE_PIM_TRUST and LSA_TRUST_ATTRIBUTE_WITHIN_FOREST bits
set will be ignored.
FreeIPA did default to LSA_TRUST_ATTRIBUTE_WITHIN_FOREST for domains
stored with trusted attributes set to 0 or missing. This was behavior in
past for Samba as well. Since FreeIPA only supports forest trust and an
external trust, in both cases we should set explicit LSA trust attribute
bits:
- LSA_TRUST_ATTRIBUTE_FOREST_TRANSITIVE for the forest trust
- LSA_TRUST_ATTRIBUTE_NON_TRANSITIVE for the external trust
Fixes: https://pagure.io/freeipa/issue/9852
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Julien Rische <jrische@redhat.com>
[MS-DRSR] 5.64 FOREST_TRUST_INFORMATION defines version of the
ForestTrustInfo structure as 1. We didn't set it so it was defaulting to
zero. Samba validates the version number and rejects ForestTrustInfo
structures from FreeIPA.
Make sure new structures are always set to version 1 and old strutures
fixed up to have version 1.
Fixes: https://pagure.io/freeipa/issue/9852
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Julien Rische <jrische@redhat.com>
When a client or server is configured with DoT, the installer
creates /etc/unbound/conf.d/zzz-ipa.conf with
tls-cert-bundle: /etc/pki/tls/certs/ca-bundle.crt
This certificate bundle does not exist any more in fedora 43+
(https://fedoraproject.org/wiki/Changes/droppingOfCertPemFile)
and the missing file prevents unbound service from starting.
The configuration should rather set
tls-system-cert: yes
in order to rely on the systemwide trust store.
The upgrade is also handled by this change.
Fixes: https://pagure.io/freeipa/issue/9838
Signed-off-by: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
During hsm validation temporary directory and files are prepared
by root in base temporary directory which can be private and
thereby not readable by anybody but root. The validation in its
turn, run some checks as pkiuser and fail to read not accessible
resources in such a case.
Hence, to make them available some shared place need to be used,
e.g. /tmp.
Fixes: https://pagure.io/freeipa/issue/9831
Signed-off-by: Stanislav Levin <slev@altlinux.org>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
The correct third party trademarks are GitHub and GitLab, and this is
the capitalization that needs to be used for documentation and messages,
when referring to each service.
Fixes: https://pagure.io/freeipa/issue/9811
Signed-off-by: Rafael Guterres Jeffman <rjeffman@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Deploying a replica with DNS support using an IPA server DNS with DoT
fails while setting up DNS over TLS. The request for the certificate for
DoT using IPA CA is done after the DNS server for the replica is configured.
The nameserver in /etc/resolv.conf has been changed to 127.0.0.1, but
unbound was not yet configured as a forwarder.
The solution is to move the cert request before the DNS server
configuration. The unbound config from the client deployment is still
working at that moment.
Fixes: https://pagure.io/freeipa/issue/9808
Signed-off-by: Thomas Woerner <twoerner@redhat.com>
Reviewed-By: Antonio Torres <antorres@redhat.com>
Unbound is by default configured to do DNSSEC validation with the validator
module. The solution is to set module-config to "iterator".
When the server is built with EDNS client subnet support this should be
changed to "subnetcache iterator" according to the unbound man page.
Fixes: https://pagure.io/freeipa/issue/9805
Signed-off-by: Thomas Woerner <twoerner@redhat.com>
Reviewed-By: Antonio Torres <antorres@redhat.com>
Reviewed-By: Alexander Bokovoy <abbra@users.noreply.github.com>
Improve how we process attributes that might include replication state
data. Previously we only cared about ";binary" but there are other
attribute options that are used in IPA. Now we completely break down the
attribute into each option and rebuild it without any repl state options
Fixes: https://pagure.io/freeipa/issue/9784
Signed-off-by: Mark Reynolds <mreynolds@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Provide the master key encryption type and the list of supported
encryption types as ipaplatform.<platform>.tasks methods. This allows
to generate the list at runtime based on the environment (e.g. FIPS) and
override the list depending of the platform.
3DES HMAC-SHA1 encryption type is now removed from supported encryption
types if it is present.
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Setting the HSM options in a pki-override-file will
result in a successful installation but it affects the
storage of the token library path. It is stored as None.
We could also instead save the provided value from the
override but an override isn't necessary in this case
since IPA provides command-line options to enable it.
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
If ipa-idrange-fix is called on a system where the server is not configured,
it crashes with a Traceback when trying to access api.env.basedn.
Check that IPA server is configured before processing further
ipatests: add test launching ipa-idrange-fix on unconfigured server
Fixes: https://pagure.io/freeipa/issue/9809
Signed-off-by: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: David Hanina <dhanina@redhat.com>
The admin must always own this name. If another entry has this
value set then remove it.
There is a uniqueness plugin for this attribute so the only two
possibilities are:
- no entry has this value set
- the admin user has this value set
- a different entry has the value set
Still, for robustness purposes, the upgrade plugin will handle
more entries.
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Provides simple warning when creating new user with uid out of
all local ranges, as this is the main culprit of breaking Kerberos, by
not generating ipantsecurityidentifier. We don't have to check for
user-mod, because modification never changes ipantsecurityidentifier.
We do not have to check groups, as groups are ignored for ipa without
AD trust. It's reasonable to revisit this in the future for group
creation and warn against groups out of ranges as well as
warn for users with groups without SID, in case AD trust is enabled.
Fixes: https://pagure.io/freeipa/issue/9781
Signed-off-by: David Hanina <dhanina@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
When ipa config-mod is called with --ipaconfigstring="", the command
fails with an InternalError.
This happens because the code added for 32bits uid did not properly
handle this case.
Same issue if ipa subid-stats is called with a null ipaconfigstring.
This commit now handles when ipaconfigstring is empty or None, and adds
a test.
Fixes: https://pagure.io/freeipa/issue/9794
Signed-off-by: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Francisco Trivino <ftrivino@redhat.com>
This has been already required for some time, just not really enforced.
Also adds few new tests, and removes test without providing rid.
Fixes: https://pagure.io/freeipa/issue/9779
Signed-off-by: David Hanina <dhanina@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
pylint 3.3+ considers the last 'return' in the flow 'useless':
contrib/perflog:89: [R1711(useless-return), parselog.display_times] Useless return at end of function or method)
ipatests/pytest_ipa/integration/tasks.py:1342: [R1711(useless-return), two_connected_topo] Useless return at end of function or method)
ipaserver/plugins/baseldap.py:575: [R1711(useless-return), host_is_master] Useless return at end of function or method)
ipaserver/install/replication.py:1425: [R1711(useless-return), ReplicationManager.initialize_replication] Useless return at end of function or method)
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
ipa-dnskeysyncd daemon relies on both OpenDNSSEC and BIND accessing the
same cryptographic token. We use SoftHSMv2 here and store token in
DNSSEC_TOKENS_DIR, defined by the IPA platform.
Configure ipa-dnskeysyncd service to update permissions of the token
files using custom systemd-tmpfiles configuration.
Extend SELinux policy to handle access to the token under a separate
file context. Both token and its pin file need to be accessed by the BIND
rndc tool.
Fixes: https://pagure.io/freeipa/issue/9696
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
For OpenSSL Provider API use we don't need 'named -E engine-name'
anymore, it has to be removed. The removal process is slightly
complicated because we need to detect '-E engine-name' and compare it
with the engine we know about (pkcs11) but if we are upgrading to the
build that supports OpenSSL Provider API, we don't know the engine name
anymore.
Fixes: https://pagure.io/freeipa/issue/9696
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
OpenSSL Engine API is deprecated and ability to compile against it is
removed in RHEL10. OpenSSL provider API is the future.
Fedora 42+ also defaults to OpenSSL provider. With pkcs11-provider, the
same PKCS#11 modules can be loaded transparently like with OpenSSL
engines. Thus, we can update configuration to use the provider API.
While Fedora 41 also defaults to OpenSSL provider, we need BIND version
that supports using OpenSSL provider API. This backport was only done in
Fedora 42.
Fixes: https://pagure.io/freeipa/issue/9696
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
When values are "normalized/converted" to a new domain the order in
which the host/release/suffix are converted matters. Replacing the
suffix first can lead to incorrect results, so convert the host/realm
before converting the suffix
Signed-off-by: Mark Reynolds <mreynolds@redhat.com>
relates: https://pagure.io/freeipa/issue/9776
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Remove replication state information (happens when LDIFs are used).
State information is written like:
attribute;adcsn=<CSN>
But we also support ";binary" which should not be removed so special
handling is needed in that case.
Signed-off-by: Mark Reynolds <mareynol@redhat.com>
Fixes: https://pagure.io/freeipa/issue/9776
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
In the wild a public CA issued a new subordinate CA certificate
with an identical subject to another, with a new private key.
This was uninstallable using ipa-cacert-manage because it would
fail with "subject public key info mismatch" during verification
because a different certificate with the same subject but
different public key was installed.
I'm not sure of the reasoning to prevent this situation but I
see it as giving users flexibility. This may be hurtful to them
but they can always remove any affected certs.
This is backwards compatible with older releases from the client
perspective. Older servers will choke on the duplicates and
won't be able to manage these.
A new serial number option is added for displaying the list of
certificates and for use when deleting one with a duplicate subject.
ipa-cacert-manage delete on systems without this patch will
successfully remove ALL of the requested certificates. There is no
way to distinguish. At least it won't break anything and the
deleted certificates can be re-added.
Fixes: https://pagure.io/freeipa/issue/9652
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
With the support of 32 bit idrange, the minvalue was set to 1
but this introduces a regression in the command ipa trust-add
as the range for AD trust is added with baserid=0
Lower the minvalue to 0
Fixes: https://pagure.io/freeipa/issue/9765
Signed-off-by: Florence Blanc-Renaud <flo@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
JSON format allows to encode integers up to 2^53-1. Linux systems allow
for 32-bit IDs. Permit setting full 32-bit uidNumber and gidNumber
through IPA API. Administrators already can set 32-bit IDs via LDAP.
ID Range also needs to permit larger sizes of RID bases. SIDGEN plugin
already treats RID bases as 1..MAX_UINT32.
Fixes: https://pagure.io/freeipa/issue/9757
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
When full 32-bit ID range usage is required, subordinate ID support have
to be disabled. However, even if ID range for subordinate IDs were to be
removed, it will be restored during the next data upgrade.
Change upgrade code to only apply subID range creation when subID
support is enabled.
Do not allow allocating subIDs if their use is disabled.
Allow full 32-bit uidNumber/gidNumber values in JSON payload.
Fixes: https://pagure.io/freeipa/issue/9757
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Disables --raw and --structured for dnsrecord-* command.
This is being shown in help for structured, as raw is implemented in
almost every command, therefore people are more likely to view
structured. Also contains tests, even though this is newly noted, this
combination has never worked in the past.
Fixes: https://pagure.io/freeipa/issue/9756
Signed-off-by: David Hanina <dhanina@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Alexander Bokovoy
Florence Blanc-Renaud
Rob Crittenden
David Hanina
Antonio Torres
Stanislav Levin
Rafael Guterres Jeffman
Thomas Woerner
Mark Reynolds
Julien Rische
Miro Hrončok
Aleksandr Sharov