]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Wed, 13 Sep 2023 22:40:11 +0000 (18:40 -0400)]
remove "flatten" migration flags
Alan T. DeKok [Wed, 13 Sep 2023 22:33:15 +0000 (18:33 -0400)]
always create Net attributes as nested
Alan T. DeKok [Wed, 13 Sep 2023 19:53:08 +0000 (15:53 -0400)]
force all nested
Alan T. DeKok [Wed, 13 Sep 2023 19:30:34 +0000 (15:30 -0400)]
ignore structural types
Alan T. DeKok [Wed, 13 Sep 2023 19:15:17 +0000 (15:15 -0400)]
don't print nested types to environment variables
Alan T. DeKok [Wed, 13 Sep 2023 18:08:00 +0000 (14:08 -0400)]
swap migration flag to always true
Alan T. DeKok [Wed, 13 Sep 2023 16:49:16 +0000 (12:49 -0400)]
add fr_pair_delete_by_da_nested()
and use it in server macros. This change allows the uses of
pair_delete_request() etc. to work properly with nested attributes
Alan T. DeKok [Wed, 13 Sep 2023 16:47:21 +0000 (12:47 -0400)]
minor cleanups for EAP attributes and rearrangements
Alan T. DeKok [Wed, 13 Sep 2023 14:03:02 +0000 (10:03 -0400)]
make EAP decode nested pairs
Alan T. DeKok [Wed, 13 Sep 2023 14:02:06 +0000 (10:02 -0400)]
Handle empty lists.
Don't recurse, as that drops a literal \0 in the middle of the
output. And also adds two spaces between the brackets.
Alan T. DeKok [Wed, 13 Sep 2023 12:09:46 +0000 (08:09 -0400)]
the EAP tests depend on libfreeradius-eap-aka-sim
not libfreeradius-eap
Alan T. DeKok [Tue, 12 Sep 2023 21:11:40 +0000 (17:11 -0400)]
explicitly encode all of the suboptions
We may want to revisit the behavior of &foo.[*]. Does it mean
only that level, or all children?
Alan T. DeKok [Tue, 12 Sep 2023 20:57:48 +0000 (16:57 -0400)]
allow EAP-AKA-SIM to encode nested pairs.
The decoder is still not done. That's next.
Alan T. DeKok [Tue, 12 Sep 2023 20:38:04 +0000 (16:38 -0400)]
fix error.
It would be nice if OSX decided to rebuild things when code changed.
Alan T. DeKok [Tue, 12 Sep 2023 20:28:36 +0000 (16:28 -0400)]
remove unused assignment
Alan T. DeKok [Tue, 12 Sep 2023 18:21:06 +0000 (14:21 -0400)]
just look at the dict, instead of calling common ancestor
because if we check for a common ancestor between a da and the
dict root, we're just checking if the da is in the dict.
Alan T. DeKok [Tue, 12 Sep 2023 18:14:21 +0000 (14:14 -0400)]
minor tweaks, and link into the main nav bar
James Jones [Fri, 8 Sep 2023 19:25:55 +0000 (14:25 -0500)]
Add coverity overview to developer documentation
James Jones [Tue, 12 Sep 2023 15:23:10 +0000 (10:23 -0500)]
Set pair list verified on initialization (true) and append (false)
CIDs: #
1544620 through #
1544624
Alan T. DeKok [Tue, 12 Sep 2023 18:02:26 +0000 (14:02 -0400)]
add and use fr_pair_dcursor_child_iter_init()
all of the encoders previously just called fr_pair_dcursor_init()
for child cursors. Which meant that the iterator could return
internal attributes.
The child cursor now uses the same iterator as the parent cursor.
This ensures that it only returns the correct encodable attributes
for this protocol.
Alan T. DeKok [Tue, 12 Sep 2023 15:00:46 +0000 (11:00 -0400)]
NDEBUG...
Alan T. DeKok [Tue, 12 Sep 2023 14:57:32 +0000 (10:57 -0400)]
add UNUSED
Alan T. DeKok [Tue, 12 Sep 2023 12:53:03 +0000 (08:53 -0400)]
fix RADIUS for nested attribute encoding
the main difference is that we fix encode_extended() to correctly
handle nesting, and update the tests. As a side effect, the
encode_extended() function now always requires nesting, and if
passed flat extended attributes, will return an encoding error.
We also fix up the fr_pair_cursor_to_network() function to remove
the flat vs nested hacks. It now always expects nesting.
We also fix up fr_struct_to_network() to always expect nesting
for the trailing TLV in a struct.
Alan T. DeKok [Mon, 11 Sep 2023 18:59:14 +0000 (14:59 -0400)]
hoist encode_cursor() function to common API
Alan T. DeKok [Mon, 11 Sep 2023 18:38:48 +0000 (14:38 -0400)]
unify function prototypes
Alan T. DeKok [Mon, 11 Sep 2023 10:31:56 +0000 (06:31 -0400)]
add RFC 7710
Alan T. DeKok [Mon, 11 Sep 2023 10:23:21 +0000 (06:23 -0400)]
trust the dcursor, and don't check for flags.internal
Alan T. DeKok [Mon, 11 Sep 2023 10:11:45 +0000 (06:11 -0400)]
allow for raw members of a struct
which can only occur when the raw VP is a direct child of the
parent struct.
Alan T. DeKok [Mon, 11 Sep 2023 09:37:09 +0000 (05:37 -0400)]
reformat
Alan T. DeKok [Mon, 11 Sep 2023 09:36:47 +0000 (05:36 -0400)]
just use tabs function instead of hand-rolling things
Alan T. DeKok [Mon, 11 Sep 2023 00:20:57 +0000 (20:20 -0400)]
comment out assertion check
because it fails for update sections, where the vp being deleted
isn't parented from the list!
If we add the migration flags
-S rewrite_update=yes -S tmpl_tokenize_all_nested=yes
then the test passes. We'll leave these checks disabled until
we either find and fix the issue (if we care) or until we enable
everything nested.
Alan T. DeKok [Sun, 10 Sep 2023 23:59:35 +0000 (19:59 -0400)]
add / clear verified flag to pair lists
so that we avoid slowdowns with repeated recursion
Alan T. DeKok [Sun, 10 Sep 2023 14:07:05 +0000 (10:07 -0400)]
return OOM on oom
Alan T. DeKok [Sun, 10 Sep 2023 13:02:01 +0000 (09:02 -0400)]
we have talloc_const_free(), so just use that
Alan T. DeKok [Sun, 10 Sep 2023 12:48:37 +0000 (08:48 -0400)]
clarify code so that we can find uses of a variable
Alan T. DeKok [Sun, 10 Sep 2023 12:43:21 +0000 (08:43 -0400)]
remove unused function
Alan T. DeKok [Sun, 10 Sep 2023 12:40:54 +0000 (08:40 -0400)]
pair_afrom_da() already copies the unknown da
Alan T. DeKok [Sat, 9 Sep 2023 20:58:16 +0000 (16:58 -0400)]
try to quiet static analyzers
Alan T. DeKok [Sat, 9 Sep 2023 15:43:56 +0000 (11:43 -0400)]
clean up in preparation for further raw cleanups
fr_dict_unknown_afrom_oid_substr() now continues from where the
previous parser left off. The raw handling is now essentially
all out of the dictionaries
Alan T. DeKok [Sat, 9 Sep 2023 11:50:02 +0000 (07:50 -0400)]
allow unknown attributes to be relative, too.
As part of a general cleanup and rearrangement, fix the code,
have less code, and make it more functional
Alan T. DeKok [Sat, 9 Sep 2023 11:29:19 +0000 (07:29 -0400)]
switch encoders to use vp->vp_type
because vp_type could be different from da->type in some cases
Alan T. DeKok [Fri, 8 Sep 2023 20:15:54 +0000 (16:15 -0400)]
fr_pair_to_unknown() -> fr_pair_raw_from_pair()
because all of the callers are creating a raw pair from malformed
data, and aren't creating an unknown pair.
Alan T. DeKok [Fri, 8 Sep 2023 20:15:25 +0000 (16:15 -0400)]
more debugging, and return OOM on OOM
Alan T. DeKok [Fri, 8 Sep 2023 20:12:55 +0000 (16:12 -0400)]
only free the the vp on the error path
if the first byte is non-zero, then the result is non-zero,
and we don't need to check it on all paths
Alan T. DeKok [Fri, 8 Sep 2023 17:25:46 +0000 (13:25 -0400)]
don't bother checking chaddr size during decode
fr_dhcpv4_ok() checks if htype==1 && hlen==6, and fails if that
isn't true.
So either that test passed, or the caller didn't call fr_dhcpv4_ok()
to verify the packet. In either case, there's no reason to do
additional validation here.
Alan T. DeKok [Fri, 8 Sep 2023 19:42:17 +0000 (15:42 -0400)]
clear the value, to be consistent with other uses of free_value
Alan T. DeKok [Fri, 8 Sep 2023 19:42:02 +0000 (15:42 -0400)]
don't clear the box, clear only the value.
Alan T. DeKok [Fri, 8 Sep 2023 19:01:47 +0000 (15:01 -0400)]
hoist TLV checks to before creating the VP
because there's no point in creating the VP and then freeing it.
Alan T. DeKok [Fri, 8 Sep 2023 14:59:48 +0000 (10:59 -0400)]
use fr_pair_raw_from_network() instead of hand-rolling it
Alan T. DeKok [Fri, 8 Sep 2023 14:24:54 +0000 (10:24 -0400)]
define and use fr_dict_unknown_copy()
because we want to get rid of fr_pair_to_unknown(). The caller
should instead have created a raw attribute
Alan T. DeKok [Fri, 8 Sep 2023 14:09:41 +0000 (10:09 -0400)]
don't need to check is_raw flag
because we now always set is_raw for unknowns, so it's enough to
check the is_unknown flag
Alan T. DeKok [Fri, 8 Sep 2023 14:06:43 +0000 (10:06 -0400)]
hoist setting of "is raw" out of protocol decoders
and into the dictionary code so it's more centralized.
Another side effect is that unknown attributes are now always
"raw". Which is likely what we should always have done anyways
Alan T. DeKok [Fri, 8 Sep 2023 13:01:19 +0000 (09:01 -0400)]
hoist "is_raw" to be more public
There was already an "is_raw" field in the unresolved sub-structure
so just hoist it out, and use it instead of da->flags.is_raw
Alan T. DeKok [Fri, 8 Sep 2023 12:59:32 +0000 (08:59 -0400)]
fix unused
Alan T. DeKok [Fri, 8 Sep 2023 12:53:24 +0000 (08:53 -0400)]
use macros / functions instead of accessing fields directly
Alan T. DeKok [Fri, 8 Sep 2023 11:47:52 +0000 (07:47 -0400)]
define and use vp_raw
In preparation for moving the "raw" flag out of the dictionaries
Alan T. DeKok [Thu, 7 Sep 2023 21:41:40 +0000 (17:41 -0400)]
fix corner case CID
1544258
Alan T. DeKok [Thu, 7 Sep 2023 20:19:27 +0000 (16:19 -0400)]
use the correct name for the parnt type
Nick Porter [Fri, 8 Sep 2023 12:24:49 +0000 (13:24 +0100)]
Simplify fr_ldap_edir_get_password
Nick Porter [Fri, 8 Sep 2023 11:25:45 +0000 (12:25 +0100)]
Return an unlang_action_t from ldap bind auth functions
Nick Porter [Thu, 7 Sep 2023 18:13:59 +0000 (19:13 +0100)]
Ensure we free memory on errors
Nick Porter [Fri, 8 Sep 2023 10:01:08 +0000 (11:01 +0100)]
Handle rcode from previous async call
Nick Porter [Fri, 8 Sep 2023 09:37:14 +0000 (10:37 +0100)]
No need for p_result in async LDAP functions
It was only set on failure, and that is handled by the
UNLANG_ACTION_FAIL return code
Nick Porter [Thu, 7 Sep 2023 17:09:55 +0000 (18:09 +0100)]
Add max_request_time timer to unit_test_module
Nick Porter [Thu, 7 Sep 2023 14:42:20 +0000 (15:42 +0100)]
Typos
Nick Porter [Thu, 7 Sep 2023 13:25:22 +0000 (14:25 +0100)]
Ensure auth bind resume functions clear up after failed requests
Nick Porter [Thu, 7 Sep 2023 13:20:10 +0000 (14:20 +0100)]
Add callback for failed ldap auth binds
Nick Porter [Thu, 7 Sep 2023 12:59:27 +0000 (13:59 +0100)]
No need for request_free callback
bind_auth_ctx is parented of the trunk request so is freed automatically
Nick Porter [Fri, 8 Sep 2023 06:59:44 +0000 (07:59 +0100)]
Typo
Alan T. DeKok [Thu, 7 Sep 2023 19:24:29 +0000 (15:24 -0400)]
always decode tacacs as nested
Alan T. DeKok [Thu, 7 Sep 2023 19:19:58 +0000 (15:19 -0400)]
allow tacacs to encode nested attributes
Alan T. DeKok [Thu, 7 Sep 2023 19:09:08 +0000 (15:09 -0400)]
typos
Alan T. DeKok [Thu, 7 Sep 2023 19:04:33 +0000 (15:04 -0400)]
it's smarter to have clear-text unit tests
instead of encrypted ones. We should also add unit tests for
encrypted packets, too
Alan T. DeKok [Thu, 7 Sep 2023 18:52:05 +0000 (14:52 -0400)]
make pair_legacy_nested=true the default
EAP-SIM and TACACS+ do not yet handle nested attributes, so they
manually set the flag to false.
Alan T. DeKok [Thu, 7 Sep 2023 18:44:42 +0000 (14:44 -0400)]
remove flatten / unflatten tests and keywords
they are no longer necessary, as we are moving to nested attributes
Alan T. DeKok [Thu, 7 Sep 2023 18:02:36 +0000 (14:02 -0400)]
don't run unbound tests on OSX
it fails in CI, and we don't care enough to fix it
Arran Cudbard-Bell [Thu, 7 Sep 2023 17:58:18 +0000 (11:58 -0600)]
json: Add json escape
Alan T. DeKok [Thu, 7 Sep 2023 17:39:28 +0000 (13:39 -0400)]
minor tweaks
Alan T. DeKok [Thu, 7 Sep 2023 17:20:17 +0000 (13:20 -0400)]
tests for nested types
Alan T. DeKok [Thu, 7 Sep 2023 17:19:55 +0000 (13:19 -0400)]
loop over nested extended types
Alan T. DeKok [Thu, 7 Sep 2023 17:19:29 +0000 (13:19 -0400)]
VSAs can only contain internal attributes and VENDORs
Alan T. DeKok [Thu, 7 Sep 2023 17:17:06 +0000 (13:17 -0400)]
the RADIUS protocol cannot encode attributes of type 'group'
Alan T. DeKok [Wed, 6 Sep 2023 21:04:27 +0000 (17:04 -0400)]
check return for CID
1544245
Realistically, unlang_xlat_yield() _always_ returns
XLAT_ACTION_YIELD, and the only reason it returns anything
is so that we can do "return unlang_xlat_yield()"
Alan T. DeKok [Wed, 6 Sep 2023 18:26:59 +0000 (14:26 -0400)]
decode TLVs nested
Alan T. DeKok [Wed, 6 Sep 2023 18:22:20 +0000 (14:22 -0400)]
nested tests for wimax
Alan T. DeKok [Wed, 6 Sep 2023 18:05:09 +0000 (14:05 -0400)]
note that giving flat lists results in us creating nested ones
Nick Porter [Wed, 6 Sep 2023 18:58:42 +0000 (19:58 +0100)]
Add 'Net.*' pairs to TACACS
Alan T. DeKok [Wed, 6 Sep 2023 16:50:14 +0000 (12:50 -0400)]
VSA types are decoded as groups
note that we will have to audit all of the "decode raw" routines.
the current behavior is for the _called_ function to create a
"raw" parent if the decode fails. If the parent is itself a group,
the called function will instead have to return an error, and then
the calling function will turn it into a raw attribute
Alan T. DeKok [Wed, 6 Sep 2023 14:09:36 +0000 (10:09 -0400)]
coalesce VENDOR attributes at the same level
Alan T. DeKok [Wed, 6 Sep 2023 12:22:51 +0000 (08:22 -0400)]
VSA types are decoded as groups
Alan T. DeKok [Wed, 6 Sep 2023 12:18:25 +0000 (08:18 -0400)]
coalesce VSAs at the same level
Nick Porter [Wed, 6 Sep 2023 11:26:52 +0000 (12:26 +0100)]
Insert new pair where we were asked to
Alan T. DeKok [Wed, 6 Sep 2023 12:07:35 +0000 (08:07 -0400)]
decode as nested
except for FR_TYPE_GROUP, which gets put into the group VP, but
the attributes are decoded from the dict root
@todo - we don't coalesce vendor / VSA data types, and we should
probably do that.
Alan T. DeKok [Wed, 6 Sep 2023 01:48:32 +0000 (21:48 -0400)]
fix group decoding
on failed decode of group, create a raw attribute instead of
returning an error.
return in the group handler, so that we don't set vp->vp_tainted
for a structural data type.
Nick Porter [Wed, 6 Sep 2023 09:40:38 +0000 (10:40 +0100)]
Don't leak memory
Arran Cudbard-Bell [Wed, 6 Sep 2023 08:55:51 +0000 (02:55 -0600)]
ldap: Quiet clang
Arran Cudbard-Bell [Wed, 6 Sep 2023 05:28:39 +0000 (23:28 -0600)]
ldap: Accept profile in either URI form or DN form
Arran Cudbard-Bell [Wed, 6 Sep 2023 04:53:00 +0000 (22:53 -0600)]
ldap: Fix auth tests
Arran Cudbard-Bell [Wed, 6 Sep 2023 04:00:06 +0000 (22:00 -0600)]
ldap: Add profile xlat to allow application of arbitrary profiles
Arran Cudbard-Bell [Wed, 6 Sep 2023 03:58:02 +0000 (21:58 -0600)]
ldap: Take a ctx in fr_lap_map_expand