]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Arran Cudbard-Bell [Sat, 31 Dec 2016 17:28:38 +0000 (12:28 -0500)]
Update contributing.md
Alan T. DeKok [Fri, 23 Dec 2016 14:19:36 +0000 (09:19 -0500)]
reject delays are from when we received the packet.
Which should be a bit better behaved if we're proxying, and the
proxy takes forever to respond. We don't add delays, we just
ensure that rejects are sent no sooner than (say) 1s after
the request was received
Alan T. DeKok [Wed, 21 Dec 2016 21:56:05 +0000 (16:56 -0500)]
more checks for is_root / vendor as per previous commit
Alan T. DeKok [Wed, 21 Dec 2016 21:45:16 +0000 (16:45 -0500)]
find_by_num is only good for top-level attributes.
If you need child, call the the other functions.
Alan T. DeKok [Wed, 21 Dec 2016 21:23:50 +0000 (16:23 -0500)]
be more aggressive about checking for user name/password
when updating maps. We don't want to catch VSAs, or other
attributes...
Alan T. DeKok [Wed, 21 Dec 2016 20:35:06 +0000 (15:35 -0500)]
more docs
It would be nice if we told people how to patch Net::RADIUS
Alan T. DeKok [Wed, 21 Dec 2016 20:18:27 +0000 (15:18 -0500)]
verify request->username && password, too
Alan T. DeKok [Wed, 21 Dec 2016 18:36:00 +0000 (13:36 -0500)]
fail if we can't start time
Alan T. DeKok [Wed, 21 Dec 2016 18:02:47 +0000 (13:02 -0500)]
simplify the time calculation to remove one branch
Alan T. DeKok [Wed, 21 Dec 2016 17:24:33 +0000 (12:24 -0500)]
added note
Arran Cudbard-Bell [Wed, 21 Dec 2016 17:22:47 +0000 (12:22 -0500)]
Fix conflicting types on FreeBSD
Alan T. DeKok [Wed, 21 Dec 2016 14:31:50 +0000 (09:31 -0500)]
remove TLS initialization. We don't need it for MD5
Alan T. DeKok [Wed, 21 Dec 2016 14:11:57 +0000 (09:11 -0500)]
Added simple RADIUS server.
Which has hand-coded check / ack routines.
Alan T. DeKok [Tue, 20 Dec 2016 20:47:50 +0000 (15:47 -0500)]
EVFILT_USER / READ / WRITE are in filter, not in flags
Alan T. DeKok [Tue, 20 Dec 2016 18:10:44 +0000 (13:10 -0500)]
more debug
Alan T. DeKok [Tue, 20 Dec 2016 18:10:28 +0000 (13:10 -0500)]
we need room for num_workers messages in the control plane, too
Alan T. DeKok [Tue, 20 Dec 2016 17:03:53 +0000 (12:03 -0500)]
travis
Alan T. DeKok [Tue, 20 Dec 2016 16:48:58 +0000 (11:48 -0500)]
remove fr_worker_control_plane()
fr_worker_channel_create() should be used instead
Alan T. DeKok [Tue, 20 Dec 2016 16:40:55 +0000 (11:40 -0500)]
Include NAKs, which are required for the channels to work
If the worker doesn't NAK a request, the master has no idea
that the request has been dropped.
This NAK could arguably be a control-plane message in the
channel, with a helper function.
Alan T. DeKok [Tue, 20 Dec 2016 16:40:21 +0000 (11:40 -0500)]
don't count RTT for NAKs
which have processing_time = 0
Alan T. DeKok [Tue, 20 Dec 2016 15:45:00 +0000 (10:45 -0500)]
use #ifdef WITH_AUTH_WINBIND
Alan T. DeKok [Tue, 20 Dec 2016 15:04:08 +0000 (10:04 -0500)]
fix condition
Herwin Weststrate [Wed, 9 Nov 2016 09:29:08 +0000 (10:29 +0100)]
Allow authentication retry in winbind
A setup with the following properties:
* Active Directory backend
* FreeRadius with eap-inner-proxy
* Windows client with single sign-on
* User using different casing in username than in backend
may result in failing connections. It looks like Windows reads the
correct username from the domain server once it has logged in, and uses
that to create the MS-CHAP2-Response attribute. The User-Name attribute
is still the one with the incorrect casing, causing the authentication
to fail.
The introduced config option kicks in after a failed authentication: it
reads the correct username from the backend, tries another
authentication, and uses the found User-Name to calculate
MS-CHAP2-Response if the second authentication works.
Matthew Newton [Tue, 20 Dec 2016 13:30:27 +0000 (13:30 +0000)]
rlm_mschap: *inst is const
Alan T. DeKok [Mon, 19 Dec 2016 22:32:49 +0000 (17:32 -0500)]
typo
Alan T. DeKok [Mon, 19 Dec 2016 22:32:11 +0000 (17:32 -0500)]
added <signal.h>
Alan T. DeKok [Mon, 19 Dec 2016 22:06:03 +0000 (17:06 -0500)]
rearrange output
Alan T. DeKok [Mon, 19 Dec 2016 22:05:28 +0000 (17:05 -0500)]
set initial RTT
Alan T. DeKok [Mon, 19 Dec 2016 22:02:08 +0000 (17:02 -0500)]
remove unused macro
Alan T. DeKok [Mon, 19 Dec 2016 21:41:40 +0000 (16:41 -0500)]
more debug stats
Alan T. DeKok [Mon, 19 Dec 2016 21:35:55 +0000 (16:35 -0500)]
use correct order for event status
Alan T. DeKok [Mon, 19 Dec 2016 21:33:59 +0000 (16:33 -0500)]
fix RTT calculation
Alan T. DeKok [Mon, 19 Dec 2016 21:18:18 +0000 (16:18 -0500)]
Added worker test
Which seems to hang often. <sigh> But recompilation makes
the hang go away!
Alan T. DeKok [Mon, 19 Dec 2016 21:18:01 +0000 (16:18 -0500)]
show NOOP, too
Alan T. DeKok [Mon, 19 Dec 2016 21:17:47 +0000 (16:17 -0500)]
More debugging
Alan T. DeKok [Mon, 19 Dec 2016 21:15:10 +0000 (16:15 -0500)]
more debugging
Alan T. DeKok [Mon, 19 Dec 2016 19:28:42 +0000 (14:28 -0500)]
count the number of requests we received
Alan T. DeKok [Mon, 19 Dec 2016 19:26:16 +0000 (14:26 -0500)]
simplify debug
Alan T. DeKok [Mon, 19 Dec 2016 19:25:41 +0000 (14:25 -0500)]
simplify debug
Alan T. DeKok [Mon, 19 Dec 2016 19:19:11 +0000 (14:19 -0500)]
add number to request
Alan T. DeKok [Mon, 19 Dec 2016 18:29:33 +0000 (13:29 -0500)]
added master ctx set/get
Alan T. DeKok [Mon, 19 Dec 2016 18:10:19 +0000 (13:10 -0500)]
Minor things missed in previous commits
Alan T. DeKok [Mon, 19 Dec 2016 17:47:02 +0000 (12:47 -0500)]
debug messages
Alan T. DeKok [Mon, 19 Dec 2016 17:46:46 +0000 (12:46 -0500)]
initialize array sizes
Alan T. DeKok [Mon, 19 Dec 2016 17:46:11 +0000 (12:46 -0500)]
break out of the loop when we've found an entry
Alan T. DeKok [Mon, 19 Dec 2016 17:19:53 +0000 (12:19 -0500)]
pass paramers to el->status() in the right order
Alan T. DeKok [Mon, 19 Dec 2016 16:46:06 +0000 (11:46 -0500)]
add worker_channel_create
and assume that decode() is given a REQUEST, instead of
allocating it itself
Alan T. DeKok [Mon, 19 Dec 2016 16:17:11 +0000 (11:17 -0500)]
clarify text
Alan T. DeKok [Mon, 19 Dec 2016 16:16:55 +0000 (11:16 -0500)]
doxygen
Alan T. DeKok [Mon, 19 Dec 2016 16:09:52 +0000 (11:09 -0500)]
added worker debug function
Alan T. DeKok [Mon, 19 Dec 2016 16:07:55 +0000 (11:07 -0500)]
Added debug function
Alan T. DeKok [Mon, 19 Dec 2016 15:34:50 +0000 (10:34 -0500)]
don't look at the other end's ack
look at our view of their ack
Alan T. DeKok [Sat, 17 Dec 2016 15:22:04 +0000 (10:22 -0500)]
typo
Alan T. DeKok [Sat, 17 Dec 2016 03:38:52 +0000 (22:38 -0500)]
<sigh> clang scan isn't quite smart enough yet
Alan T. DeKok [Sat, 17 Dec 2016 01:54:36 +0000 (20:54 -0500)]
see if this shuts up travis
Arran Cudbard-Bell [Sat, 17 Dec 2016 00:45:53 +0000 (19:45 -0500)]
Exit with an error code reflecting the signal
Arran Cudbard-Bell [Sat, 17 Dec 2016 00:34:19 +0000 (19:34 -0500)]
set CC_HINT(nonnull) on all cursor functions
mark hot functions as hot.
_fr_cursor_init in particular gets flagged by profiling programs as consuming lots of CPU time.
Arran Cudbard-Bell [Sat, 17 Dec 2016 00:13:47 +0000 (19:13 -0500)]
Various doxygen and cursor code fixes
Arran Cudbard-Bell [Fri, 16 Dec 2016 22:32:54 +0000 (17:32 -0500)]
Add new generic cursor functions
Arran Cudbard-Bell [Fri, 16 Dec 2016 22:14:44 +0000 (17:14 -0500)]
s/cursor/pair_cursor/g
It'll be changed back in a future commit... maybe
Alan T. DeKok [Fri, 16 Dec 2016 18:49:37 +0000 (13:49 -0500)]
travis
Alan T. DeKok [Fri, 16 Dec 2016 18:48:22 +0000 (13:48 -0500)]
allocate reply message in worker
Alan T. DeKok [Fri, 16 Dec 2016 18:29:22 +0000 (13:29 -0500)]
add worker context to channel
Because the network signals the worker that a channel is ready,
but the worker needs additional information around that channel
And it's inefficient for the worker to root through it's list
of channels, trying to find a context
Alan T. DeKok [Fri, 16 Dec 2016 16:36:53 +0000 (11:36 -0500)]
set channel to be unused in close
Alan T. DeKok [Fri, 16 Dec 2016 16:36:01 +0000 (11:36 -0500)]
rename function to be more consistent
Alan T. DeKok [Fri, 16 Dec 2016 16:26:40 +0000 (11:26 -0500)]
remove old todo which has since been done
Alan T. DeKok [Fri, 16 Dec 2016 16:26:16 +0000 (11:26 -0500)]
use talloc_pooled_object
Alan T. DeKok [Fri, 16 Dec 2016 15:48:47 +0000 (10:48 -0500)]
Service the control plane on EVFILT_USER signal
which means that the worker should now (mostly) work with new
channels and closing old ones.
Alan T. DeKok [Fri, 16 Dec 2016 14:46:14 +0000 (09:46 -0500)]
add evfilt_user callback, and glue it into the event loop
Alan T. DeKok [Fri, 16 Dec 2016 14:39:52 +0000 (09:39 -0500)]
allocate channel array
Alan T. DeKok [Thu, 15 Dec 2016 15:39:03 +0000 (10:39 -0500)]
docs
Alan T. DeKok [Thu, 15 Dec 2016 15:38:03 +0000 (10:38 -0500)]
yielded requests go to the head of the list
Alan T. DeKok [Thu, 15 Dec 2016 15:18:15 +0000 (10:18 -0500)]
add get control plane AQ
Alan T. DeKok [Thu, 15 Dec 2016 00:52:44 +0000 (19:52 -0500)]
doxygen
Alan T. DeKok [Wed, 14 Dec 2016 20:43:03 +0000 (15:43 -0500)]
add comments explaining functional requirements
Alan T. DeKok [Wed, 14 Dec 2016 20:33:15 +0000 (15:33 -0500)]
more cleanups.
The only place to decode packets is in the main loop.
We have a heap for runnable requests, ordered by (priority, time)
We have one dlist for requests, insert into head, cleanup by tail
Alan T. DeKok [Wed, 14 Dec 2016 20:27:45 +0000 (15:27 -0500)]
added FR_DLIST_INSERT_HEAD
Alan T. DeKok [Wed, 14 Dec 2016 19:52:23 +0000 (14:52 -0500)]
resumable requests have priority over runnable ones.
And don't insert them back into the runnable list.
Alan T. DeKok [Wed, 14 Dec 2016 19:39:33 +0000 (14:39 -0500)]
clean up worker.
Move "to_decode", etc. to struct containing heap + dlist,
so that the "pop" cane be done by priority, but "get oldest"
is done by time order.
Add template unlange_resume() function so that we know what
it's supposed to do.
modify transport APIs a bit based on current design.
Alan T. DeKok [Wed, 14 Dec 2016 18:41:05 +0000 (13:41 -0500)]
Added FR_INSERT_TAIL_PTR
Alan T. DeKok [Wed, 14 Dec 2016 16:09:29 +0000 (11:09 -0500)]
added FR_DLIST_TAIL to get the tail of the list
Arran Cudbard-Bell [Tue, 13 Dec 2016 15:36:03 +0000 (10:36 -0500)]
Fix asserts in ippool_tool
Arran Cudbard-Bell [Tue, 13 Dec 2016 14:12:47 +0000 (09:12 -0500)]
Package client
Alan T. DeKok [Mon, 12 Dec 2016 20:18:05 +0000 (15:18 -0500)]
fix API
Alan T. DeKok [Mon, 12 Dec 2016 20:12:53 +0000 (15:12 -0500)]
clarify names for channel ends
Alan T. DeKok [Mon, 12 Dec 2016 20:01:52 +0000 (15:01 -0500)]
fix return code
Alan T. DeKok [Mon, 12 Dec 2016 19:43:37 +0000 (14:43 -0500)]
travis fixes
Alan T. DeKok [Mon, 12 Dec 2016 19:30:14 +0000 (14:30 -0500)]
move to control-plane signaling for channels
Alan T. DeKok [Mon, 12 Dec 2016 18:34:09 +0000 (13:34 -0500)]
more debug messages.
Should also make num_messages configurable
Alan T. DeKok [Mon, 12 Dec 2016 18:33:39 +0000 (13:33 -0500)]
add debug messages
Alan T. DeKok [Mon, 12 Dec 2016 17:54:12 +0000 (12:54 -0500)]
suppress more signals from the worker, too
Alan T. DeKok [Mon, 12 Dec 2016 17:42:20 +0000 (12:42 -0500)]
Correct logic for master -> worker signals
Which results in the suppression of many more signals
Alan T. DeKok [Mon, 12 Dec 2016 16:20:47 +0000 (11:20 -0500)]
increase size of control-plane queue based on max_outstanding
Alan T. DeKok [Mon, 12 Dec 2016 15:30:24 +0000 (10:30 -0500)]
Expose control-plane push/pop
Alan T. DeKok [Sun, 11 Dec 2016 14:38:46 +0000 (09:38 -0500)]
count resignals
Alan T. DeKok [Sun, 11 Dec 2016 14:30:06 +0000 (09:30 -0500)]
make fr_channel_service_aq() public
Alan T. DeKok [Sun, 11 Dec 2016 14:29:54 +0000 (09:29 -0500)]
remove bad fprintf
Alan T. DeKok [Sun, 11 Dec 2016 14:29:40 +0000 (09:29 -0500)]
use correct ACK
Alan T. DeKok [Sun, 11 Dec 2016 02:39:40 +0000 (21:39 -0500)]
unify public and internal signals
Alan T. DeKok [Sun, 11 Dec 2016 02:35:36 +0000 (21:35 -0500)]
create fr_channel_service_aq()
in preparation for control-plane messages