]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Sat, 24 Aug 2013 14:05:01 +0000 (10:05 -0400)]
If there's no packet, don't add module_failure_msg. Closes #407
Alan T. DeKok [Fri, 23 Aug 2013 19:37:13 +0000 (15:37 -0400)]
Don't set home server to NULL. It's needed for other things
Alan T. DeKok [Fri, 23 Aug 2013 13:06:46 +0000 (09:06 -0400)]
Fix a typo
Alan T. DeKok [Fri, 23 Aug 2013 12:46:24 +0000 (08:46 -0400)]
Whoops... handle MOD_CASE, too
Alan T. DeKok [Fri, 23 Aug 2013 02:32:58 +0000 (22:32 -0400)]
Rewrite the modcall MOD_* handlers. Updates #404.
The old code wasn't worth debugging. With a few hours work,
the new code is simpler, more understandable, and avoids the
bug entirely.
Alan T. DeKok [Tue, 20 Aug 2013 14:21:37 +0000 (10:21 -0400)]
Clean up "close socket" routines.
We use the new rbtree DeleteOrder when walking over the packet list.
When the "eol tcp/proxy listener" callback is run, it can return
"please delete this node", instead of calling rbtree_delete.
Doing that allows the walker to be aware of deletions, unlike
before.
This turned out to make the code simpler, and the whole event fd
handling has become more robust.
Arran Cudbard-Bell [Tue, 20 Aug 2013 13:30:04 +0000 (14:30 +0100)]
Port new logging macros to release branch
Arran Cudbard-Bell [Thu, 15 Aug 2013 17:16:33 +0000 (18:16 +0100)]
don't segv printing interfaceid
Arran Cudbard-Bell [Fri, 16 Aug 2013 17:47:54 +0000 (18:47 +0100)]
Don't segv if asked to print a null value
Arran Cudbard-Bell [Tue, 20 Aug 2013 13:02:30 +0000 (14:02 +0100)]
Fix segv in rlm_expiration
vp_strvalue no longer available for none string attributes
Alan T. DeKok [Fri, 16 Aug 2013 13:51:55 +0000 (09:51 -0400)]
Simplify "proxy anew" code
Now that the packet list re-uses Ids less often
Alan T. DeKok [Fri, 16 Aug 2013 13:48:24 +0000 (09:48 -0400)]
Simplify ID allocation so that we don't loop over all IDs
Instead, we allocate a 256-entry array of random IDs (all unique)
and gradually use them. When the array is empty, we re-fill it.
Alan T. DeKok [Fri, 16 Aug 2013 12:59:52 +0000 (08:59 -0400)]
Unify error logic
Alan T. DeKok [Fri, 16 Aug 2013 03:01:17 +0000 (23:01 -0400)]
Build the rbmonkey test tool
Alan T. DeKok [Fri, 16 Aug 2013 03:01:06 +0000 (23:01 -0400)]
Formatting and warning fixes
skids [Tue, 25 Jun 2013 22:52:15 +0000 (18:52 -0400)]
A test file for libradius rbtree ops
Verifies red/black counts to all child nodes after adds/deletes
Tests DeleteOrder; code provided in comments for vanilla rbtree_delete.
Build system magic necessary for this is beyond me.
It's a start at least
skids [Tue, 25 Jun 2013 22:47:14 +0000 (18:47 -0400)]
Add a batch deletion mode to rbtree_walk
Walks the tree calling callbacks as per InOrder
Allows callback to ask for deletion of records after evaluating criteria
All callbacks and deletions done while lock is held
skids [Thu, 20 Jun 2013 20:03:20 +0000 (16:03 -0400)]
Add rbtree_callbydata for sane threadsafe/garbage-collected operations
When working with an rbtree which is exposed to multiple threads,
it is not safe to do much of anything with data retrieved by functions
such as rbtree_finddata, other than to feed it directly to rbtree_delete.
This is because once rbtree_finddata has returned, another thread may
obtain a pointer to that data (and hence may begin mangling non-key
material such as container_of or any child allocs.) This is especially
true for rbtrees that have a freeNode garbage collection routine defined.
This function allows safe operations while the rbtree lock is still
held. It also allows for a conditional delete operation based on
criteria which may only be safe to ascertain while the lock is held.
In addition to short operations, it could be used, with due care, to
trylock a more granular lock associated with the key before deleting
the key or before operating with it outside of the rbtree lock.
Alan T. DeKok [Thu, 15 Aug 2013 13:12:33 +0000 (09:12 -0400)]
Remove redundant headers
Alan T. DeKok [Thu, 15 Aug 2013 13:12:17 +0000 (09:12 -0400)]
Remove redundant headers
Alan T. DeKok [Wed, 14 Aug 2013 16:29:59 +0000 (12:29 -0400)]
packet list id alloc/free now inserts / yanks the packet, too
Alan T. DeKok [Wed, 14 Aug 2013 16:21:33 +0000 (12:21 -0400)]
Made more functions return bool
Alan T. DeKok [Wed, 14 Aug 2013 16:17:00 +0000 (12:17 -0400)]
Remove talloc asserts
some of the time, the packet is on the stack, and not talloc'd
Alan T. DeKok [Wed, 14 Aug 2013 14:59:05 +0000 (10:59 -0400)]
Return codes are boolean
Alan T. DeKok [Wed, 14 Aug 2013 14:40:57 +0000 (10:40 -0400)]
Missed in earlier commit
Alan T. DeKok [Wed, 14 Aug 2013 14:23:10 +0000 (10:23 -0400)]
Use correct arguments to talloc destructors
Alan T. DeKok [Wed, 14 Aug 2013 14:14:55 +0000 (10:14 -0400)]
Check return code of packet_list_yank
Alan T. DeKok [Wed, 14 Aug 2013 14:14:32 +0000 (10:14 -0400)]
Made packet_list_yank return a bool for success/failure
Alan T. DeKok [Wed, 14 Aug 2013 14:09:13 +0000 (10:09 -0400)]
Added VERIFY_PACKET, too
Alan T. DeKok [Tue, 13 Aug 2013 13:06:15 +0000 (09:06 -0400)]
Fix more issues for new v3 API
Olivier Beytrison [Tue, 13 Aug 2013 09:56:44 +0000 (11:56 +0200)]
Correct call to paircreate in radeapclient
Alan T. DeKok [Mon, 12 Aug 2013 19:39:30 +0000 (15:39 -0400)]
Fix horrific spacing issue.
Alan T. DeKok [Mon, 12 Aug 2013 14:21:55 +0000 (10:21 -0400)]
Actually use virtual server when proxying to it. Closes #375
Alan T. DeKok [Mon, 12 Aug 2013 13:16:47 +0000 (09:16 -0400)]
free -> talloc_free
Alan T. DeKok [Sun, 11 Aug 2013 14:39:09 +0000 (10:39 -0400)]
Only print foreach close brace once. Closes #377
Alan T. DeKok [Fri, 9 Aug 2013 15:13:36 +0000 (11:13 -0400)]
Remove from the proxy list if we can't insert it
Alan T. DeKok [Fri, 9 Aug 2013 13:26:00 +0000 (09:26 -0400)]
Don't smash passwords with newly alloc'd zero array
Arran Cudbard-Bell [Wed, 7 Aug 2013 16:57:31 +0000 (17:57 +0100)]
Don't iterate over head attribute twice Fixes #401
Alan T. DeKok [Wed, 7 Aug 2013 01:05:07 +0000 (21:05 -0400)]
Updates from latest doc
Arran Cudbard-Bell [Tue, 6 Aug 2013 12:46:01 +0000 (13:46 +0100)]
Typo in default ldap config
Arran Cudbard-Bell [Tue, 6 Aug 2013 11:27:16 +0000 (12:27 +0100)]
Fix logic error in rlm_ldap lockout
Alan T. DeKok [Tue, 6 Aug 2013 20:45:50 +0000 (16:45 -0400)]
Revert "Attribute 0 is forbidden"
This reverts commit
d5b6011a6ffb27c48a12b694bb9cfcfc0f7c5466 .
Alan T. DeKok [Tue, 6 Aug 2013 20:45:35 +0000 (16:45 -0400)]
Oh well
Revert "Attribute codes of zero are forbidden"
This reverts commit
12da0d625f0bd5f35c6887e9050eeb189329ca27 .
Alan T. DeKok [Tue, 6 Aug 2013 20:44:42 +0000 (16:44 -0400)]
Attribute 0 is forbidden
Alan T. DeKok [Tue, 6 Aug 2013 20:12:58 +0000 (16:12 -0400)]
Attribute codes of zero are forbidden
Alan T. DeKok [Tue, 6 Aug 2013 20:12:44 +0000 (16:12 -0400)]
These guys are winners.
Please enter the commit message for your changes. Lines starting
Alan T. DeKok [Tue, 6 Aug 2013 20:10:19 +0000 (16:10 -0400)]
Broadsoft dictionary
Alan T. DeKok [Tue, 6 Aug 2013 15:31:30 +0000 (11:31 -0400)]
Doing nothing is a NOOP
Alan T. DeKok [Tue, 6 Aug 2013 14:25:11 +0000 (10:25 -0400)]
Accounting-Requests don't need User-Name
Alan T. DeKok [Fri, 2 Aug 2013 12:28:17 +0000 (08:28 -0400)]
Remove unused assignments
Arran Cudbard-Bell [Fri, 2 Aug 2013 10:30:59 +0000 (11:30 +0100)]
Otherattr should return false by default, so that the check attribute is searched for
Arran Cudbard-Bell [Tue, 30 Jul 2013 22:53:58 +0000 (23:53 +0100)]
Use the same const order everywhere
Arran Cudbard-Bell [Tue, 30 Jul 2013 11:43:56 +0000 (12:43 +0100)]
re-enable option to disable shared libraries (which causes binaries to be statically linked against internal libraries)
Arran Cudbard-Bell [Tue, 30 Jul 2013 11:35:28 +0000 (12:35 +0100)]
Typo in libtool.mk
Arran Cudbard-Bell [Tue, 30 Jul 2013 10:43:41 +0000 (11:43 +0100)]
If were linking to the server library, we don't need to redefine module_failure_msg
Arran Cudbard-Bell [Tue, 30 Jul 2013 08:28:03 +0000 (09:28 +0100)]
Reverse VALUE_PAIR const order
Arran Cudbard-Bell [Tue, 30 Jul 2013 08:26:05 +0000 (09:26 +0100)]
Reverse DICT_ATTR const order
Olivier Beytrison [Mon, 29 Jul 2013 15:31:42 +0000 (17:31 +0200)]
Move paircompare to DICT_ATTR
All modules registering paircompare callback have been updated.
For every callback which didn't use the *req list, the first_only has been set to true
Arran Cudbard-Bell [Mon, 29 Jul 2013 09:17:33 +0000 (10:17 +0100)]
Add assert to catch mismatched attributes
Alan T. DeKok [Fri, 26 Jul 2013 21:33:19 +0000 (17:33 -0400)]
Set fake packet timestamp from request packet timestamp
Alan T. DeKok [Fri, 26 Jul 2013 21:29:03 +0000 (17:29 -0400)]
Move setting of packet->timestamp
Arran Cudbard-Bell [Fri, 26 Jul 2013 20:42:01 +0000 (21:42 +0100)]
More configuration sanity checks for rlm_ldap
Arran Cudbard-Bell [Fri, 26 Jul 2013 19:19:26 +0000 (20:19 +0100)]
Only skip on vendor if were searching for a callback
Arran Cudbard-Bell [Fri, 26 Jul 2013 12:45:48 +0000 (13:45 +0100)]
Later versions of GCC complain that the value of ftruncate is unused
Why? Because GCC is fucking retarded.
Arran Cudbard-Bell [Fri, 26 Jul 2013 12:18:16 +0000 (13:18 +0100)]
Use result buffer of localtime_r, it's less confusing
Arran Cudbard-Bell [Fri, 26 Jul 2013 11:45:53 +0000 (12:45 +0100)]
Don't crash if we have a zero length expansion made up of one component
Arran Cudbard-Bell [Fri, 26 Jul 2013 11:38:58 +0000 (12:38 +0100)]
Add assert to catch incorrect attributes
Conflicts:
src/modules/rlm_expr/paircmp.c
Arran Cudbard-Bell [Fri, 26 Jul 2013 11:21:39 +0000 (12:21 +0100)]
Only clear matches on new match so nested regex comparisons work correctly
Also add some useful debug output
Arran Cudbard-Bell [Fri, 26 Jul 2013 10:43:56 +0000 (11:43 +0100)]
We only support paircmp for NON vendor attributes #381
Arran Cudbard-Bell [Fri, 26 Jul 2013 09:28:00 +0000 (10:28 +0100)]
Fix argument names in radius_callback_compare
Alan T. DeKok [Thu, 25 Jul 2013 23:26:05 +0000 (19:26 -0400)]
Fix installation rules for sites / modules. Closes #360
The sites-enabled/* and mods-enabled/* were still listed as
installation targets. So they were INSTALLed, even when the
"create soft link" rule was if'd out.
The solution is to add both the target and the rule only when
necessary
Alan T. DeKok [Thu, 25 Jul 2013 14:41:51 +0000 (10:41 -0400)]
Be more specific with install rules. Closes #360
Arran Cudbard-Bell [Thu, 25 Jul 2013 21:32:06 +0000 (22:32 +0100)]
fix decoding of client ethernet address (by mino98)
Arran Cudbard-Bell [Thu, 25 Jul 2013 15:46:37 +0000 (16:46 +0100)]
Fix deadlock in eap_handler_free, and also simplify eap handler free
Arran Cudbard-Bell [Thu, 25 Jul 2013 12:58:50 +0000 (13:58 +0100)]
Closes #391
Arran Cudbard-Bell [Thu, 25 Jul 2013 11:25:45 +0000 (12:25 +0100)]
Fix for opaque data double free
Alan T. DeKok [Thu, 25 Jul 2013 00:52:15 +0000 (20:52 -0400)]
Set reply timestamp always
Arran Cudbard-Bell [Wed, 24 Jul 2013 13:30:59 +0000 (14:30 +0100)]
Unify regex subcapture group processing
Arran Cudbard-Bell [Wed, 24 Jul 2013 11:15:10 +0000 (12:15 +0100)]
Buffer no longer used
Arran Cudbard-Bell [Wed, 24 Jul 2013 11:06:20 +0000 (12:06 +0100)]
Fix all other instances of sizeof(vp->vp_strvalue)
Arran Cudbard-Bell [Wed, 24 Jul 2013 10:46:21 +0000 (11:46 +0100)]
If pairparsevalue gets called on a regex pair, we DON'T attempt to parse the value, and instead copy the string value over verbatim.
Arran Cudbard-Bell [Wed, 24 Jul 2013 10:42:20 +0000 (11:42 +0100)]
char buffer[sizeof(check->vp_strvalue)];
Guessing we don't want a 4 or 8 byte buffer...
Arran Cudbard-Bell [Wed, 24 Jul 2013 10:40:46 +0000 (11:40 +0100)]
Skip the entry if expansion fails for one of the check items
Arran Cudbard-Bell [Wed, 24 Jul 2013 10:40:03 +0000 (11:40 +0100)]
Fixup pairparsevalue signature (should return bool)
Fix places which assumed because it returned an int, it would return a negative integer to indicate failure.
Arran Cudbard-Bell [Wed, 24 Jul 2013 10:00:14 +0000 (11:00 +0100)]
xlat check items too
Arran Cudbard-Bell [Tue, 23 Jul 2013 21:21:50 +0000 (22:21 +0100)]
Fix segv in rlm_ldap_search
Arran Cudbard-Bell [Tue, 23 Jul 2013 21:05:59 +0000 (22:05 +0100)]
s/[ \t]+$//g - config and docs
Arran Cudbard-Bell [Tue, 23 Jul 2013 20:17:49 +0000 (21:17 +0100)]
Fix flower boxes
Arran Cudbard-Bell [Tue, 23 Jul 2013 20:16:02 +0000 (21:16 +0100)]
Skip membership checks if no membership attribute is configured
Arran Cudbard-Bell [Tue, 23 Jul 2013 19:57:51 +0000 (20:57 +0100)]
rlm_python is NOT python3.x ready so don't search for version 3. Fix python configure script to work with 3.x
Arran Cudbard-Bell [Tue, 23 Jul 2013 19:18:18 +0000 (20:18 +0100)]
Fix segv in rlm_ldap_search
Arran Cudbard-Bell [Tue, 23 Jul 2013 17:32:39 +0000 (18:32 +0100)]
Add assert to cacheable grou code
Arran Cudbard-Bell [Tue, 23 Jul 2013 20:20:12 +0000 (21:20 +0100)]
Always get the result when were searching so we can count entries, and check for errors counting results.
Arran Cudbard-Bell [Tue, 23 Jul 2013 11:42:41 +0000 (12:42 +0100)]
Return if we can't find the VP CID
1047333
Alan T. DeKok [Tue, 23 Jul 2013 11:49:49 +0000 (07:49 -0400)]
Fixes for coverity
Alan T. DeKok [Tue, 23 Jul 2013 11:44:22 +0000 (07:44 -0400)]
Print correct value
Arran Cudbard-Bell [Tue, 23 Jul 2013 11:33:17 +0000 (12:33 +0100)]
Arran Cudbard-Bell [Tue, 23 Jul 2013 10:12:57 +0000 (11:12 +0100)]
We have to free packets properly with rad_free instead of just free
Arran Cudbard-Bell [Tue, 23 Jul 2013 09:51:50 +0000 (10:51 +0100)]
We *MUST NOT* free the data pointer explicitly. Anything that allocates and expects it to be freed MUST allocate from the context of the packet.
This caused abort() in radsniff
Alan Buxey [Mon, 22 Jul 2013 21:45:28 +0000 (22:45 +0100)]
all IPv4 examples should now be using RFC5737 documentation space
also fixed 2 'tunnelled' - we now have tunneled as standard
(code/config options have that spelling)