ipasam already implemented retrieval of groups for MS-SAMR calls.
However, it did not have implementation of a group retrieval for the
path of lookup_name() function in Samba. The lookup_name() is used in
many places in smbd and winbindd.
With this change it will be possible to resolve IPA groups in Windows UI
(Security tab) and console (net localgroup ...). When Global Catalog
service is enabled, it will be possible to search for those groups as
well.
In Active Directory, security groups can be domain, domain local, local
and so on. In IPA, only domain groups exposed through ipasam because
SID generation plugin only supports adding SIDs to POSIX groups and
users. Thus, non-POSIX groups are not going to have SIDs associated and
will not be visible in both UNIX and Windows environments.
Group retrieval in Samba is implemented as a mapping between NT and
POSIX groups. IPA doesn't have explicit mapping tables. Instead, any
POSIX group in IPA that has a SID associated with it is considered a
domain group for Samba.
Finally, additional ACI is required to ensure attributes looked up by
ipasam are always readable by the trust agents.
Fixes: https://pagure.io/freeipa/issue/8660
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
When MS-PAC structure is created, we get passed the time of
authentication from KDC. Use this to record logon time in MS-PAC
structure.
Set allow password change time to the last password change. We need to
refer to the actual password policy here in future.
Also use INT64_MAX to represent the resulting value for logoff
and kickoff times according to MS-PAC 2.6.
Fixes: https://pagure.io/freeipa/issue/8659
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
ipadb_get_principal() allocates client_actual. Call
ipadb_free_principal to release it.
Rather than spreading the free() amongst the code introduce
done as a target to match behavior in similar functions.
Discovered by coverity.
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Discovered by coverity
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
The purpose is to set a default if the RID doesn't match
expectations.
Discovered by coverity
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Discovered by coverity.
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
krb5 uses the negative part of krb5_timestamp to store time values
after 2038:
https://k5wiki.kerberos.org/wiki/Projects/Timestamps_after_2038
In other words, krb5 uses krb5_timestamp (signed int) with
unsigned arithmetic for expanding the timestamp's upper bound.
This commit:
- adds some helper functions for working with krb5_timestamp as
unsigned (actually copied from
https://github.com/krb5/krb5/blob/master/src/include/k5-int.h)
- replaces operations with krb5_timestamp's by these new functions
Fixes: https://pagure.io/freeipa/issue/8028
Signed-off-by: Slava Aseev <ptrnine@altlinux.org>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
In case we've got a principal name as '*', we don't need to specify
the principal itself, use pre-defined filter for a wild-card search.
Previously, we had to escape the '*' as specifying it with an explicit
matching rule would have violated RFC 4515 section 3. However, since we
don't really need to specify a different matching rule for a wild-card
search, we can remove this part completely.
Use this change as an opportunity to simplify the code and reduce
number of duplicated filter constants -- if extra filter is NULL, we can
simply pass "" and use _EXTRA filter constants to format the final
filter.
Fixes: https://pagure.io/freeipa/issue/8624
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
Python code does detection of the system encoding based on the locale
settings. On RHEL 8.4 development images we somehow get LANG=en_US which
defaults to iso8859-1 _inside_ the systemd-started service, even though
the whole environment defaults to LANG=en_US.UTF-8.
When instrumented with ExecStartPre=/usr/bin/locale, the following
output can be seen:
locale[45481]: LANG=en_US
locale[45481]: LC_CTYPE="en_US"
locale[45481]: LC_NUMERIC="en_US"
locale[45481]: LC_TIME="en_US"
locale[45481]: LC_COLLATE="en_US"
locale[45481]: LC_MONETARY="en_US"
locale[45481]: LC_MESSAGES="en_US"
locale[45481]: LC_PAPER="en_US"
locale[45481]: LC_NAME="en_US"
locale[45481]: LC_ADDRESS="en_US"
locale[45481]: LC_TELEPHONE="en_US"
locale[45481]: LC_MEASUREMENT="en_US"
locale[45481]: LC_IDENTIFICATION="en_US"
locale[45481]: LC_ALL=
ipactl[45483]: Unexpected error
ipactl[45483]: SystemEncodingError: System encoding must be UTF-8, 'iso8859-1' is not supported. Set LC_ALL="C.UTF-8", or LC_ALL="" and LC_CTYPE="C.UTF-8".
systemd[1]: ipa.service: Main process exited, code=exited, status=1/FAILURE
Set the environment to explicit LC_ALL=C.UTF-8 to please the Python
code. FreeIPA server side only cares about actual encoding, not the
language itself. We already use LC_ALL=C.UTF-8 in httpd service snippet.
Fixes: https://pagure.io/freeipa/issue/8617
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Thomas Woerner <twoerner@redhat.com>
Commit 8b6d1ab854 added support for
subordinate UPN suffixes but missed the case where subordinate UPN is a
subdomain of the forest root domain and not mentioned in the UPN
suffixes list.
Correct this situation by applying the same check to the trusted domain
name as well.
Fixes: https://pagure.io/freeipa/issue/8554
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
In file included from /usr/include/string.h:519,
from /usr/include/lber.h:30,
from /usr/include/ldap.h:30,
from internal.h:27,
from main.c:31:
In function 'strncpy',
inlined from 'main' at main.c:237:5:
/usr/include/bits/string_fortified.h:91:10: warning: 'strncpy' specified bound 255 equals destination size [-Wstringop-truncation]
91 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fixes: https://pagure.io/freeipa/issue/8585
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
In file included from /usr/include/string.h:519,
from /usr/include/lber.h:30,
from /usr/include/ldap.h:30,
from ipa_sam.c:12:
In function 'strncpy',
inlined from 'save_sid_to_secret' at ipa_sam.c:4478:2,
inlined from 'pdb_init_ipasam' at ipa_sam.c:4985:12:
/usr/include/bits/string_fortified.h:91:10: warning: 'strncpy' specified bound 255 equals destination size [-Wstringop-truncation]
91 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fixes: https://pagure.io/freeipa/issue/8585
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
We use string_to_sid() from internal Samba libraries, so we have to link
to it properly.
In addition, size_t is (long unsigned int), just cast to (int) in
asprintf.
Fixes: https://pagure.io/freeipa/issue/8585
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
In file included from /usr/include/string.h:519,
from /usr/include/lber.h:30,
from /usr/include/ldap.h:30,
from ipa_kdb.h:37,
from ipa_kdb_mspac.c:26:
In function 'strncpy',
inlined from 'get_server_netbios_name' at ipa_kdb_mspac.c:2358:5,
inlined from 'ipadb_reinit_mspac' at ipa_kdb_mspac.c:2813:39:
/usr/include/bits/string_fortified.h:91:10: warning: 'strncpy' specified bound 255 equals destination size [-Wstringop-truncation]
91 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fixes: https://pagure.io/freeipa/issue/8585
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Jitter is always enabled, so there is no additional configuration.
An earlier version of this patch was authored by Becky Shanley.
Fixes: https://pagure.io/freeipa/issue/8010
Signed-off-by: Robbie Harwood <rharwood@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
This fixes the case where an account is locked on one or more servers
and the password is reset by an administrator. The account would
remain locked on those servers for the duration of the lockout.
This is done by setting krbLastAdminUnlock to the current date and
time. The lockout plugin will see this and unlock the account. Since
the value should be replicated along with the password any server
that has the new password will also be unlocked.
This does incur an additional attribute that must be replicated,
whether it is needed or not, but since lockout is computed
per-server this is the only guaranteed way to be sure that the
account will be unlocked everywhere.
My original thought was to grab password replication events and detect
whether the user was locked out and unlock them. On any given server
you can only know if the user is locked out on that server by
computing it. Doing this would require generalizing the lockout code
so it could be computed on password change. krbLastFailedAuth could
be wiped which would unlock the account on that master (the attribute
is not replicated by default).
So it is complexity vs additional replication. Assuming that admin
reset is relatively rare let's start with that. This doesn't lock
us into this solution for the future.
We could set this attribute on user-driven password changes as
well but the original ask and my thinking are that if you forgot
your password and got locked out, how can you change it yourself?
Upon reflection I guess a user could fat-finger it a bunch of times
against one IPA server then have a revelation and log in against a
different server. So they would still be locked out for the duration
on the first one. I'm not sure the extra replication is worth it for
user-generated password changes or that users would be saavy enough
to try another server for the change.
https://pagure.io/freeipa/issue/8551
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
When initializing UPN suffixes, we calculate their sizes and didn't use
the right variable to allocate their size. This affects us if there are
more than one UPN suffix available for a trust due to memory corruption
while filling in sizes.
Add unit test for multiple UPN suffixes.
Fixes: https://pagure.io/freeipa/issue/8566
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
[MS-ADTS] 6.1.6.9.3.2 requires msDS-TrustForestTrustInfo attribute of
trusted domain information in Active Directory to conform certain rules.
One side-effect of those rules is that list of UPN suffixes reported
through the netr_DsRGetForestTrustInformation function is dynamically
filtered to deduplicate subordinate suffixes.
It means that if list of UPN suffixes contains the following top level
names (TLNs):
fabrikam.com
sub.fabrikam.com
then netr_DsRGetForestTrustInformation would only return 'fabrikam.com'
as the TLN, fully filtering 'sub.fabrikam.com'.
IPA KDB driver used exact comparison of the UPN suffixes so any
subordinate had to be specified exactly.
Modify logic so that if exact check does not succeed, we validate a
realm to test being a subordinate of the known UPN suffixes. The
subordinate check is done by making sure UPN suffix is at the end of the
test realm and is immediately preceded with a dot.
Because the function to check suffixes potentially called for every
Kerberos principal, precalculate and cache length for each UPN suffix at
the time we retrieve the list of them.
Fixes: https://pagure.io/freeipa/issue/8554
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Robbie Harwood <rharwood@redhat.com>
ipa_sam was modifying the buffer returned by ipa_gethostfqdn().
Subsequent calls to ipa_gethostfqdn() returned the corrupt data,
causing other operations to fail.
Update ipa_sam to copy the string and modify the copy. Also
document this characteristic of ipa_gethostfqdn() and explain that
callers must not modify the returned data.
Part of: https://pagure.io/freeipa/issue/8501
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
ipa_gethostfqdn() now returns a pointer to a statically allocated buffer
or NULL in case of an error. The caller no longer has to supply a
correctly allocated buffer.
Rename IPA_HOST_HOST to_LEN IPA_HOST_FQDN_LEN and use IPA_HOST_FQDN_LEN
wherever code copies a hostname supplied from ipa_gethostfqdn().
Clarify that MAXHOSTNAMELEN and MAXHOSTFQDNLEN are different things.
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
ipa_kdb and ipa-join now use ipa_gethostfqdn() instead of uname()'s nodename.
The code for hostname in ipa-join is simplified. Now the hostname is
auto-detected and verified in main(). All sub functions can now use the
hostname without additional checks. This removes a bunch of strdup(),
NULL checks, and free() calls.
Fixes: https://pagure.io/freeipa/issue/8501
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
FreeIPA's Python and C code used different approaches to get the FQDN of
the host. Some places assumed that gethostname() returns a FQDN. Other
code paths used glibc's resolver to resolve the current node name to a
FQDN.
Python code now uses the ipalib.constants.FQDN where a fully qualified
domain name is expected. The variable is initialized only once and avoids
potential DNS lookups.
C code uses a new helper function ipa_gethostfqdn() in util package. The
function implements similar logic as gethostfqdn() except it uses more
modern getaddrinfo(). The result is cached as well.
Fixes: https://pagure.io/freeipa/issue/8501
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Fraser Tweedale <ftweedal@redhat.com>
Enable checking:
maxrepeat - reject passwrods which contain more than N consecutive
characters.
maxsequence - rejected passwords which contain character sequences
(abcde).
dictcheck - check passwords using cracklib
usercheck - check whether the password contains the user name.
The class checking provided by libpwpolicy is not used because this
overlaps with the existing IPA checking. This includes the options
dcredit, ucredit, lcredit, ocredit, minclass and maxclassrepeat.
The pwquality min length is fixed at 6 so if there is a conflict between
the system policy and pwquality log that length is enforced at 6.
https://pagure.io/freeipa/issue/6964https://pagure.io/freeipa/issue/5948https://pagure.io/freeipa/issue/2445https://pagure.io/freeipa/issue/298
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
cast const error message to non-const char*. I tried to make errMesg a
const char* but it gets passed down to slapi_send_ldap_result() which
accepts a char*.
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
kadmin.local getprincs command results in passing '*' as a principal to
KDB driver function that looks up the principals.
The whole filter looks like this
(&(|
(objectclass=krbprincipalaux)
(objectclass=krbprincipal)
(objectclass=ipakrbprincipal))
(|(ipakrbprincipalalias=*)
(krbprincipalname:caseIgnoreIA5Match:=*)))
There are two parts of the LDAP filter we use to look up principals, the
part with 'krbprincipalname' uses extensible filter syntax of RFC 4515
section 3:
extensible = ( attr [dnattrs]
[matchingrule] COLON EQUALS assertionvalue )
/ ( [dnattrs]
matchingrule COLON EQUALS assertionvalue )
In case we've got a principal name as '*' we have to follow RFC 4515
section 3 and reencode it using <valueencoding> rule from RFC 4511
section 4.1.6 but only to the part of the filter that does use assertion
value.
Fixes: https://pagure.io/freeipa/issue/8490
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Previously, the logging level of StreamHandler for ipa-dnskeysyncd
was restricted to INFO via `standard_logging_setup(verbose=False)`.
Thus, it was impossible to get messages having lower level.
This also sets the loglevel for ipa-dnskeysyncd to DEBUG for
troubleshooting.
Fixes: https://pagure.io/freeipa/issue/8094
Signed-off-by: Stanislav Levin <slev@altlinux.org>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Unit tests for ipa-extdom-extop plugin use nss_files.so.2 module to test the
functionality instead of relying on SSSD API or nss_sss.so.2 module. The latter
two cannot be used in build environment.
nss_files.so.2 always tries to open /etc/passwd and /etc/group. In past, we
overloaded 'fopen()' to change the path to opened file but this stops working
after glibc consolidate file opening in nss_files with the code starting at
https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=299210c1fa67e2dfb564475986fce11cd33db9ad,
this method is not usable anymore and builds against glibc 2.31.9000+ fail in
cmocka unit test execution in Rawhide.
Apply an alternative approach that uses a new user namespace to unshare the
test from its parent and chroot to the test data where expected /etc/passwd and
/etc/group are provided. This method works only on Linux, thus only run the
unit test on Linux.
In case unshare() or chroot() fail, we have to skip tests that use
nss_files.so.2.
Fixes: https://pagure.io/freeipa/issue/8437
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Some places have to use the old name because it's part of the stable API
or stable LDAP attributes.
See: https://tools.ietf.org/id/draft-knodel-terminology-01.html
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abbra@users.noreply.github.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
Use OpenSSL's HMAC API instead of NSS.
Fixes: Fixes: https://pagure.io/freeipa/issue/6857
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Refactoring of the get_principal() code in commit
b5876f30d4 broke handling of enterprise
principal lookup for AS request (kinit -E user@ipa.test@IPA.TEST).
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Rob Crittenden <rcritten@redhat.com>
"Kerberos principal expiration" is set in UTC and when server is in
different timezone, the time difference between timezone is respected by
the IPA server/client for Kerberos authentication.
The problem is due to mktime() assuming default time zone but since we
parse the time using Zulu (UTC+0) timezone, mktime() forces current time
zone offset added.
The method is using mktime() and comparing to the current time obtained
with time(NULL). According to its man page, mktime is considering the
time as local time:
The mktime() function converts a broken-down time structure, expressed
as local time, to calendar time representation.
Instead mktime() we should use timegm(). The problem is that it is
non-standard GNU extension and it is recommended (in the man page for
timegm(3)) to avoid its use. An alternative is to set TZ=UTC, call
mktime(), unset TZ, but since we are running in a multi-threaded
environment this is problematic.
On the other hand, we already rely on GNU extensions and enable them
with -D_DEFAULT_SOURCE=1, so use of timegm() is enabled already.
The fix, therefore, is to use timegm() instead of mktime() in
daemons/ipa-slapi-plugins/ipa-pwd-extop/prepost.c in two places where we
first do 'strptime()' with Zulu time zone (in ipapwd_pre_bind() and
ipapwd_write_krb_keys()).
Fixes: https://pagure.io/freeipa/issue/8362
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-by: Simo Sorce <simo@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Restructure logic of ipadb_get_principal() to separate retrieval of a
principal by a name and by an alias. Separate enterprise principal name
type processing into a helper function to be able to reuse it for own
aliases.
Unify code in client referrals part to do the same and use krb5 API to
deal with principals rather than parsing strings. The end result is the
same but we follow common rules in MIT Kerberos to process principals.
An enterprise principal is typically "name@SOMEREALM@REALM", but any
principal might be parsed as enterprise principal, so we could get
"name@REALM" marked as such. When unparsing the enterprise principal,
re-parse it again with default realm values, to get our realm
normalization.
This behavior would fix situations when GSSAPI calls are operating on a
non-qualified principal name that was imported as a
GSS_KRB5_NT_ENTERPRISE_NAME when calling gss_import_name().
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Signed-off-by: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
For Kerberos principal lookup we always need to check whether principal
is from our realm. Keep the reference to our realm TGS handy to avoid
memory allocations on every lookup.
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Somehow, we weren't adding primary group of the user to the list of
groups in the PAC Logon Info structure.
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Previously, FreeIPA only allowed to issue PAC record in a ticket
for the following principal types:
- for IPA users
- for a host principal of one of IPA masters
- for a cifs/ or HTTP/ service on one of IPA masters
To allow S4U2Self operations over trust to AD, an impersonating service
must have PAC record in its TGT to be able to ask AD DCs for a S4U2Self
ticket. It means any IPA service performing S4U2Self would need to have
PAC record and the constraints above prevent it from doing so.
However, depending on whether the service or host principal belongs to
one of IPA masters, we need to set proper primary RID to 516 (domain
controllers) or 515 (domain computers).
Fixes: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Depending on whether identity of a principal was asserted by the KDC or
by a service doing protocol transition (S4U2Self), AD DCs add a
special extra SID to a PAC record:
- S-1-18-1 is a SID for an Authentication Authority Asserted Identity
- S-1-18-2 is a SID for a Service Asserted Identity
This behavior is governed by [MS-SFU] 3.2.5.1.2 "KDC replies with Service
Ticket".
In order to add an asserted identity SID, we need to pass down the
client flags as set by the KDC and check for a protocol transition bit.
Fixes: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Implement minimal server referrals support for enterprise principals as
defined in RFC 6806.
Use krb5_pac_verify_ext() and krb5_pac_sign_ext() to support cross-realm
S4U extensions. We have to verify/sign PAC and take the realm into
account for S4U in these cases.
The use of extended functions require krb5 1.17+.
For PAC verification, we have to filter existing PAC CLIENT-INFO
structure in cross-realm S4U case because otherwise old CLIENT-INFO
would change the PAC principal due to adding or ommiting the realm in
transition. Since a new PAC CLIENT-INFO will be provided by
k5_insert_client_info() anyway, we can filter it in all cases.
Generate PAC only for the first S4U2Self request to the client realm
(client != NULL). Otherwise, use the PAC from the cross-realm ticket.
The latter PAC belongs to the impersonated user.
Foreign (inner) principal look up in non-AS request returns
KRB5_KDB_NOENTRY.
Finally, in PAC signing we have to take the realm into account as well
for S4U2Self cross-realm operation. This does not work when compiling
against krb5 1.17 at the moment because sign_authdata() callback does
not know whether we are dealing with an issuing referral or not. In 1.18
a KDC will set a special client flag to signify this when asking KDB
driver to sign a PAC record.
Fixes: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Signed-off-by: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
Helper utility to investigate PAC content of users in trusted
environments. Supports direct ticket acquisition and S4U2Self protocol
transition.
1. Direct ticket acquisition
In direct ticket acquisition mode the utility first does one of the
following actions:
- obtain a TGT ticket for a user principal using supplied password
- import existing TGT from a default credentials cache
Once a user TGT is available, the utility will attempt to acquire a service
ticket to a service which key is specified in a keytab (default or
passed with --keytab option) and simulate establishing context to the
service application.
If establishing context succeeds, MS-PAC content of the service ticket
will be printed out.
2. S4U2Self protocol transition
In protocol transition case a service application obtains own TGT using
a key from the keytab and then requests a service ticket to itself in
the name of the user principal, performing S4U2Self request.
If accepting this service ticket succeeds, MS-PAC content of the service
ticket will be printed out.
If KDC does not support or rejects issuing MS-PAC record for a user, an
error message 'KDC has no support for padata type' will be printed.
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Signed-off-by: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
UPN_DNS_INFO structure contains the client's user principal name (UPN)
and a fully qualified domain name. It is used to provide the UPN and the
FQDN that corresponds to the client of the ticket.
The structure is defined in MS-PAC section 2.10. MS-KILE specification
says in the section 3.3.5.6.4.5 that KDCs should return this buffer. It
further clarifies in section 3.3.5.2 that if the user account object has no
userPrincipalName attribute, UPN_DNS_INFO should be constructed by
concatenating user name, the "@" symbol, and the DNS name of the domain.
IPA users don't really have userPrincipalName attribute. Instead, we
always construct their account names in LOGON Info3 structure by
unparsing the canonical principal name without realm, meaning that user
principal can be recovered by concatenating the account name and the
realm (domain).
Unless the account name and unparsed client principal name are different
or the primary Info3 gid (group RID) is the one for machine accounts,
mark the UPN as constructed.
Related: https://pagure.io/freeipa/issue/8319
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Isaac Boukris <iboukris@redhat.com>
Reviewed-By: Florence Blanc-Renaud <flo@redhat.com>
BER structure representing a string might not have termination '\0'
character, thus we should use length-bound functions to operate on it.
Memory handling of LDAP values was leaving previous vals over iteration.
Also, when freeing vals, we need to explicitly set it to NULL.
Fixes: https://pagure.io/freeipa/issue/8291
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-by: Simo Sorce <ssorce@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Simo Sorce <ssorce@redhat.com>
Related: https://pagure.io/freeipa/issue/8291
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Simo Sorce <ssorce@redhat.com>
Require OpenDNSSEC 2.1.6-5 with fix for RHBZ#1825812 (DAC override AVC)
Allow ipa-dnskeysyncd to connect to enforcer.sock (ipa_dnskey_t write
opendnssec_var_run_t and connectto opendnssec_t). The
opendnssec_stream_connect interface is available since 2016.
Change the owner of the ipa-ods-exporter socket to ODS_USER:ODS_GROUP.
The ipa-ods-exporter service already runs as ODS_USER.
Fixes: https://pagure.io/freeipa/issue/8283
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Florence Blanc-Renaud <frenaud@redhat.com>
Also add runstatedir autoconf var. IPA requires autoconf 2.59. The
variable will be available with autoconf 2.70.
Fixes: https://pagure.io/freeipa/issue/8272
Signed-off-by: Christian Heimes <cheimes@redhat.com>
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
NIST SP 800-63-3B sets a recommendation to have password length upper bound limited in A.2:
https://pages.nist.gov/800-63-3/sp800-63b.html#appA
Users should be encouraged to make their passwords as lengthy as they
want, within reason. Since the size of a hashed password is independent
of its length, there is no reason not to permit the use of lengthy
passwords (or pass phrases) if the user wishes. Extremely long passwords
(perhaps megabytes in length) could conceivably require excessive
processing time to hash, so it is reasonable to have some limit.
FreeIPA already applied 256 characters limit for non-random passwords
set through ipa-getkeytab tool. The limit was not, however, enforced in
other places.
MIT Kerberos limits the length of the password to 1024 characters in its
tools. However, these tools (kpasswd and 'cpw' command of kadmin) do not
differentiate between a password larger than 1024 and a password of 1024
characters. As a result, longer passwords are silently cut off.
To prevent silent cut off for user passwords, use limit of 1000
characters.
Thus, this patch enforces common limit of 1000 characters everywhere:
- LDAP-based password changes
- LDAP password change control
- LDAP ADD and MOD operations on clear-text userPassword
- Keytab setting with ipa-getkeytab
- Kerberos password setting and changing
Fixes: https://pagure.io/freeipa/issue/8268
Signed-off-by: Alexander Bokovoy <abokovoy@redhat.com>
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
Reviewed-by: Simo Sorce <ssorce@redhat.com>
Reviewed-By: Simo Sorce <ssorce@redhat.com>
While other password policies were properly ignored the password
history was always being saved if the global history size was
non-zero.
Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
Reviewed-By: Christian Heimes <cheimes@redhat.com>
