The corresponding open is done as part of initializing a connection_struct
object, where we chdir() and stat() the root path of the share. The stat() in
vfs_fruit causes an expensive metadata request on the path which triggers an
internal open of a pathref handle. Note that this only affects servers that have
fruit:metadata = netatalk set, which is the default unfortunately.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Fri Apr 7 21:12:21 UTC 2023 on atb-devel-224
Ralph Boehme [Wed, 5 Apr 2023 09:03:52 +0000 (11:03 +0200)]
smbd: Prevent creation of vetoed files
The problem is when checking for vetoed names on the last path component in
openat_pathref_fsp_case_insensitive() we return
NT_STATUS_OBJECT_NAME_NOT_FOUND. The in the caller
filename_convert_dirfsp_nosymlink() this is treated as the "file creation case"
causing filename_convert_dirfsp_nosymlink() to return NT_STATUS_OK.
In order to correctly distinguish between the cases
1) file doesn't exist, we may be creating it, return
2) a vetoed a file
we need 2) to return a more specific error to
filename_convert_dirfsp_nosymlink(). I've chosen NT_STATUS_OBJECT_NAME_INVALID
which gets mapped to the appropriate errror NT_STATUS_OBJECT_PATH_NOT_FOUND or
NT_STATUS_OBJECT_NAME_NOT_FOUND depending on which path component was vetoed.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Thu Apr 6 23:03:50 UTC 2023 on atb-devel-224
Dmitry Antipov [Fri, 17 Mar 2023 10:06:52 +0000 (13:06 +0300)]
lib:registry: drop unused argument of reg_open_remote()
Drop 'struct auth_session_info *' argument of reg_open_remote()
which is actually unused (NULL passed by all of the callers).
Signed-off-by: Dmitry Antipov <dantipov@cloudlinux.com> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu Apr 6 16:46:03 UTC 2023 on atb-devel-224
Signed-off-by: Andreas Schneider <asn@samba.org> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu Apr 6 14:47:20 UTC 2023 on atb-devel-224
Andrew Bartlett [Thu, 6 Apr 2023 00:28:12 +0000 (12:28 +1200)]
selftest: Use "debug syslog format = always" in selftest
Some of the most difficult to debug issues in Samba development are around
timing, so this changes our default logging format in the selftest system
to include a high-resolution timestamp to help correlate bad events with
what else is going on at the same time.
This fits in well with the timestamps already logged into st/subunit
and may assist with correlation.
Signed-off-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Thu Apr 6 13:44:47 UTC 2023 on atb-devel-224
Signed-off-by: Rob van der Linde <rob@catalyst.net.nz> Reviewed-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Thu Apr 6 01:33:05 UTC 2023 on atb-devel-224
dsdb/tests: Add test for modification of unicodePwd over a cleartext/signed connection
This demonstrates that the server did not detect CVE-2023-0922
Signed-off-by: Andrew Bartlett <abartlet@samba.org> Signed-off-by: Rob van der Linde <rob@catalyst.net.nz> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Signed-off-by: Rob van der Linde <rob@catalyst.net.nz> Reviewed-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Andrew Bartlett [Wed, 5 Apr 2023 20:59:17 +0000 (08:59 +1200)]
dsdb/tests: Double number of expressions in large_ldap.py ldap_timeout test
By slowing the filter down more this makes the test reliable on the
autobuild host.
This is not a long-term solution, but is a quick tweak that can be done
today to address current issues with getting commits past the host-based
(compared with cloud-based) autobuild.
Signed-off-by: Rob van der Linde <rob@catalyst.net.nz> Signed-off-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Wed Apr 5 03:08:51 UTC 2023 on atb-devel-224
Joseph Sutton [Sun, 8 Jan 2023 22:22:34 +0000 (11:22 +1300)]
CVE-2023-0225 s4-acl: Don't return early if dNSHostName element has no values
This early return would mistakenly allow an unprivileged user to delete
the dNSHostName attribute by making an LDAP modify request with no
values. We should no longer allow this.
Add or replace operations with no values and no privileges are
disallowed.
Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz> Reviewed-by: Joseph Sutton <josephsutton@catalyst.net.nz> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Joseph Sutton [Thu, 23 Feb 2023 21:03:25 +0000 (10:03 +1300)]
CVE-2023-0614 s4-dsdb: Treat confidential attributes as unindexed
In the unlikely case that someone adds a confidential indexed attribute
to the schema, LDAP search expressions on that attribute could disclose
information via timing differences. Let's not use the index for searches
on confidential attributes.
Joseph Sutton [Fri, 3 Mar 2023 04:35:55 +0000 (17:35 +1300)]
CVE-2023-0614 ldb: Filter on search base before redacting message
Redaction may be expensive if we end up needing to fetch a security
descriptor to verify rights to an attribute. Checking the search scope
is probably cheaper, so do that first.
Joseph Sutton [Fri, 3 Mar 2023 04:34:29 +0000 (17:34 +1300)]
CVE-2023-0614 ldb: Prevent disclosure of confidential attributes
Add a hook, acl_redact_msg_for_filter(), in the aclread module, that
marks inaccessible any message elements used by an LDAP search filter
that the user has no right to access. Make the various ldb_match_*()
functions check whether message elements are accessible, and refuse to
match any that are not. Remaining message elements, not mentioned in the
search filter, are checked in aclread_callback(), and any inaccessible
elements are removed at this point.
Certain attributes, namely objectClass, distinguishedName, name, and
objectGUID, are always present, and hence the presence of said
attributes is always allowed to be checked in a search filter. This
corresponds with the behaviour of Windows.
Further, we unconditionally allow the attributes isDeleted and
isRecycled in a check for presence or equality. Windows is not known to
make this special exception, but it seems mostly harmless, and should
mitigate the performance impact on searches made by the show_deleted
module.
As a result of all these changes, our behaviour regarding confidential
attributes happens to match Windows more closely. For the test in
confidential_attr.py, we can now model our attribute handling with
DC_MODE_RETURN_ALL, which corresponds to the behaviour exhibited by
Windows.
Pair-Programmed-With: Andrew Bartlett <abartlet@samba.org>
Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz> Signed-off-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Joseph Sutton [Thu, 26 Jan 2023 19:32:41 +0000 (08:32 +1300)]
CVE-2023-0614 tests/krb5: Add test for confidential attributes timing differences
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15270 Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
which overwrote the parse tree on an existing ldb request with a fixed
up tree. This became a problem if a module performed another search with
that same request structure, as extended_dn_in would try to fix up the
already-modified tree for a second time. The fixed-up tree element now
having an extended DN, it would fall foul of the ldb_dn_match_allowed()
check in extended_dn_filter_callback(), and be replaced with an
ALWAYS_FALSE match rule. In practice this meant that <GUID={}> searches
would only work for one search in an ldb request, and fail for
subsequent ones.
Fix this by creating a new request with the modified tree, and leaving
the original request unmodified.
Joseph Sutton [Sun, 26 Feb 2023 21:31:52 +0000 (10:31 +1300)]
CVE-2023-0614 ldb: Make use of ldb_filter_attrs_in_place()
Change all uses of ldb_kv_filter_attrs() to use
ldb_filter_attrs_in_place() instead. This function does less work than
its predecessor, and no longer requires the allocation of a second ldb
message. Some of the work is able to be split out into separate
functions that each accomplish a single task, with a purpose to make the
code clearer.
Joseph Sutton [Fri, 3 Mar 2023 04:30:19 +0000 (17:30 +1300)]
CVE-2023-0614 ldb: Make ldb_filter_attrs_in_place() work in place
ldb_filter_attrs() previously did too much. Now its replacement,
ldb_filter_attrs_in_place(), only does the actual filtering, while
taking ownership of each element's values is handled in a separate
function, ldb_msg_elements_take_ownership().
Also, ldb_filter_attrs_in_place() no longer adds the distinguishedName
to the message if it is missing. That is handled in another function,
ldb_msg_add_distinguished_name().
As we're now modifying the original message rather than copying it into
a new one, we no longer need the filtered_msg parameter.
We adapt a test, based on ldb_filter_attrs_test, to exercise the new
function.
Joseph Sutton [Fri, 3 Mar 2023 04:29:03 +0000 (17:29 +1300)]
CVE-2023-0614 ldb: Add function to filter message in place
At present this function is an exact duplicate of ldb_filter_attrs(),
but in the next commit we shall modify it to work in place, without the
need for the allocation of a second message.
The test is a near duplicate of the existing test for
ldb_filter_attrs().
Joseph Sutton [Fri, 3 Mar 2023 04:23:42 +0000 (17:23 +1300)]
CVE-2023-0614 ldb: Add function to take ownership of an ldb message
Many places in Samba depend upon various components of an ldb message
being talloc allocated, and hence able to be used as talloc contexts.
The elements and values of an unpacked ldb message point to unowned data
inside the memory-mapped database, and this function ensures that such
messages have talloc ownership of said elements and values.
Joseph Sutton [Tue, 14 Feb 2023 23:34:51 +0000 (12:34 +1300)]
CVE-2023-0614 ldb:tests: Ensure ldb_val data is zero-terminated
If the value of an ldb message element is not zero-terminated, calling
ldb_msg_find_attr_as_string() will cause the function to read off the
end of the buffer in an attempt to verify that the value is
zero-terminated. This can cause unexpected behaviour and make the test
randomly fail.
To avoid this, we must have a terminating null byte that is *not*
counted as part of the length, and so we must calculate the length with
strlen() rather than sizeof.
projects / thirdparty / samba.git / log
