]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Wed, 18 Oct 2017 18:08:21 +0000 (14:08 -0400)]
assert that packets which have replies were actually sent
i.e. once we have a reply, the packet is no longer sent,
it's done.
Alan T. DeKok [Tue, 17 Oct 2017 20:48:50 +0000 (16:48 -0400)]
set work_io_instance=NULL in file, not worker
because many things can cause the file to be deleted...
Alan T. DeKok [Tue, 17 Oct 2017 20:16:02 +0000 (16:16 -0400)]
more cleanups on error, and deletion of detail.work
Alan T. DeKok [Tue, 17 Oct 2017 20:00:20 +0000 (16:00 -0400)]
alloc a worker and run it
Alan T. DeKok [Tue, 17 Oct 2017 19:58:57 +0000 (15:58 -0400)]
add decode & write
which bounces through to proto_detail_work, which does the
actual work.
We should probably have a way to pass the fr_listen_t to proto_detail,
so it can use it's app_io, instead of the one in the proto_detail
instance
Alan T. DeKok [Tue, 17 Oct 2017 19:57:52 +0000 (15:57 -0400)]
free the worker if we're closing the FD
and if the flag tells us to
Alan T. DeKok [Tue, 17 Oct 2017 19:41:09 +0000 (15:41 -0400)]
minor cleanups
add "self" for linking issues.
parent is not "const"
move fr_schedule_t to proto_detail_t, not proto_detail_work_t
and initialize the entries
Alan T. DeKok [Tue, 17 Oct 2017 19:40:46 +0000 (15:40 -0400)]
clean up to allow proto_detail_file to call us
Alan T. DeKok [Tue, 17 Oct 2017 18:18:37 +0000 (14:18 -0400)]
minor cleanups
filename.work is no longer mandatory. If it isn't defined,
we'll just create it.
If detail.work does exist, watch for it's deletion
Alan T. DeKok [Tue, 17 Oct 2017 18:18:22 +0000 (14:18 -0400)]
more than one NOTE can be returned
Alan T. DeKok [Tue, 17 Oct 2017 17:48:37 +0000 (13:48 -0400)]
open RDWR if tracking, else RDONLY
Alan T. DeKok [Tue, 17 Oct 2017 17:32:07 +0000 (13:32 -0400)]
clear out el->events[] for an FD when the event is deleted
Alan T. DeKok [Tue, 17 Oct 2017 14:49:10 +0000 (10:49 -0400)]
use -, not a-b
Alan T. DeKok [Tue, 17 Oct 2017 13:12:51 +0000 (09:12 -0400)]
don't use t->pending, just rely on heap size
if there are two things showing the same information, it's possible
for them to get out of sync. So... one way to rule them all.
Alan T. DeKok [Tue, 17 Oct 2017 13:08:28 +0000 (09:08 -0400)]
set t->pending only if there's pending packets
Alan T. DeKok [Mon, 16 Oct 2017 19:31:00 +0000 (15:31 -0400)]
handle 'close' as a separate function
so we can catch it, and inform our parent instance that it
needs to rescan the directory
Alan T. DeKok [Mon, 16 Oct 2017 19:30:40 +0000 (15:30 -0400)]
instance isn't 'const' in close
Alan T. DeKok [Mon, 16 Oct 2017 19:18:27 +0000 (15:18 -0400)]
auto-create "work" configuration
to simplify the users life
Alan T. DeKok [Mon, 16 Oct 2017 19:18:16 +0000 (15:18 -0400)]
note recent changes
Alan T. DeKok [Mon, 16 Oct 2017 18:55:24 +0000 (14:55 -0400)]
/ may occur in dictionary names, too
so we can't put it into the comments in the file
Alan T. DeKok [Mon, 16 Oct 2017 18:53:39 +0000 (14:53 -0400)]
cast to long double
Alan T. DeKok [Mon, 16 Oct 2017 18:53:06 +0000 (14:53 -0400)]
%f is double
Alan T. DeKok [Mon, 16 Oct 2017 18:52:08 +0000 (14:52 -0400)]
cast to double
Alan T. DeKok [Mon, 16 Oct 2017 18:44:38 +0000 (14:44 -0400)]
simplify the translation rules
Alan T. DeKok [Mon, 16 Oct 2017 18:30:07 +0000 (14:30 -0400)]
port relay fixes from 3.0
Alan T. DeKok [Fri, 13 Oct 2017 17:17:03 +0000 (13:17 -0400)]
sort by FD first, not filter.
there are really only 2 filters used, but multiple FDs
Alan T. DeKok [Fri, 13 Oct 2017 13:58:56 +0000 (09:58 -0400)]
don't free children in parallel state destructor
they're already either parented from the request, or linked
to the request via request_data_add. And will therefore be
freed as necessary.
Plus, if they're detached, we definitely DON'T want to free them
when the parallel section exits
Alan T. DeKok [Fri, 13 Oct 2017 13:53:13 +0000 (09:53 -0400)]
use correct enum value
Alan T. DeKok [Fri, 13 Oct 2017 13:49:58 +0000 (09:49 -0400)]
parallel clones by default
but can be empty via 'parallel empty'
which likely needs a better name
Alan T. DeKok [Fri, 13 Oct 2017 13:21:09 +0000 (09:21 -0400)]
free children on DONE signal
Alan T. DeKok [Fri, 13 Oct 2017 11:21:45 +0000 (07:21 -0400)]
delete the reply after receiving it
Alan T. DeKok [Fri, 13 Oct 2017 02:12:13 +0000 (22:12 -0400)]
connection may be NULL
Alan T. DeKok [Fri, 13 Oct 2017 01:40:25 +0000 (21:40 -0400)]
remove unused variable
Alan T. DeKok [Fri, 13 Oct 2017 01:39:46 +0000 (21:39 -0400)]
dereference correct variable
Alan T. DeKok [Thu, 12 Oct 2017 19:15:18 +0000 (15:15 -0400)]
make ourselves active if we read and there's pending packets
this is likely imperfect, but it's not entirely wrong
Alan T. DeKok [Thu, 12 Oct 2017 19:04:12 +0000 (15:04 -0400)]
double-check idle handling
Alan T. DeKok [Thu, 12 Oct 2017 18:57:34 +0000 (14:57 -0400)]
more cleanups and sanity checks
Alan T. DeKok [Thu, 12 Oct 2017 18:30:41 +0000 (14:30 -0400)]
go over status-server retransmission checks
Alan T. DeKok [Thu, 12 Oct 2017 18:12:36 +0000 (14:12 -0400)]
go over zombie timeout checks
Alan T. DeKok [Wed, 11 Oct 2017 20:10:04 +0000 (16:10 -0400)]
and more cleanups
the basics work. But (again) the corner cases need to be checked.
Alan T. DeKok [Wed, 11 Oct 2017 19:39:15 +0000 (15:39 -0400)]
more cleanups
Alan T. DeKok [Wed, 11 Oct 2017 19:16:40 +0000 (15:16 -0400)]
start of cleanup and change to better state transitions
Alan DeKok [Wed, 11 Oct 2017 17:58:50 +0000 (13:58 -0400)]
Merge pull request #2091 from bmork/unisphere-vsas-for-v4.0.x
add new Juniper/Unisphere/ERX VSAs
Alan T. DeKok [Wed, 11 Oct 2017 16:08:25 +0000 (12:08 -0400)]
remove active_requests
we can calculate it from the queued / sent heaps
Alan T. DeKok [Wed, 11 Oct 2017 13:25:12 +0000 (09:25 -0400)]
travis fix
Alan T. DeKok [Wed, 11 Oct 2017 12:50:53 +0000 (08:50 -0400)]
clean up connection state on failure
Alan T. DeKok [Tue, 10 Oct 2017 19:26:40 +0000 (15:26 -0400)]
always add Event-Timestampt to status packets
and always update it when sending those packets
Alan T. DeKok [Tue, 10 Oct 2017 19:20:29 +0000 (15:20 -0400)]
don't update Acct-Delay-Time for status check packets
Alan T. DeKok [Tue, 10 Oct 2017 19:18:04 +0000 (15:18 -0400)]
enforce limits on retransmissions
and make status check packets use the same retransmission path
as normal packets
Alan T. DeKok [Tue, 10 Oct 2017 18:59:46 +0000 (14:59 -0400)]
use local retry variable
Alan T. DeKok [Tue, 10 Oct 2017 18:54:44 +0000 (14:54 -0400)]
we can send reply to network after max_request_time
Alan T. DeKok [Tue, 10 Oct 2017 18:53:39 +0000 (14:53 -0400)]
do proper state changes on retransmit
Bjørn Mork [Wed, 11 Oct 2017 12:01:26 +0000 (14:01 +0200)]
add new Juniper/Unisphere/ERX VSAs
Importing new attributes from
https://www.juniper.net/documentation/software/junos/junos161/radius-dictionary/unisphereDictionary_for_JUNOS_v16-1.dct
using the exact attribute names Juniper use, following the precedence set
by the original conversion from dictionary.erx to dictionary.unisphere
Yes, Juniper *do* use four different attributes prefixes here:
Unisphere, Pcp (sic), Jnpr and ERX(!)
Quite impressive. And quite a mess. But at least it's not *our* mess
if we just got with their names...
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Arran Cudbard-Bell [Wed, 11 Oct 2017 07:21:05 +0000 (13:21 +0600)]
Return -1 from fr_event_filter_insert if fr_event_build_evset fails (instead of going into an infinite loop)
Alan T. DeKok [Tue, 10 Oct 2017 18:43:12 +0000 (14:43 -0400)]
track active packets in a connection
Alan T. DeKok [Tue, 10 Oct 2017 18:26:17 +0000 (14:26 -0400)]
add explicit STATE to udp_request_t
and other minor fixes:
- use allowed_replies[code], not code2rcode[code]
- ensure that all packets are in the right list
- switch off of state to do packet list enqueue / dequeue
Alan T. DeKok [Tue, 10 Oct 2017 15:26:54 +0000 (11:26 -0400)]
re-initialize fields after re-open
Alan T. DeKok [Tue, 10 Oct 2017 14:59:30 +0000 (10:59 -0400)]
don't decrement num_requests on status_server
Alan T. DeKok [Tue, 10 Oct 2017 13:33:08 +0000 (09:33 -0400)]
more assertions and cleanups
Alan T. DeKok [Tue, 10 Oct 2017 13:12:50 +0000 (09:12 -0400)]
clean up more request stuff, and add more assertions
Arran Cudbard-Bell [Tue, 10 Oct 2017 11:40:50 +0000 (17:40 +0600)]
Table used outside of when WITH_EVENT_DEBUG is defined now
Arran Cudbard-Bell [Tue, 10 Oct 2017 11:39:41 +0000 (17:39 +0600)]
Use kevent_filter_table
Alan T. DeKok [Tue, 10 Oct 2017 11:13:48 +0000 (07:13 -0400)]
ifdef out array for EVENT_DEBUG
Alan T. DeKok [Mon, 9 Oct 2017 13:01:41 +0000 (09:01 -0400)]
%f is double, so an explicit conversion is useful
as a hint to the compiler that we know what we're doing
Arran Cudbard-Bell [Tue, 10 Oct 2017 10:51:15 +0000 (16:51 +0600)]
EVFILT_EXCEPT not defined by libkqueue
Arran Cudbard-Bell [Tue, 10 Oct 2017 10:34:10 +0000 (16:34 +0600)]
add extra logging to event.c
Arran Cudbard-Bell [Tue, 10 Oct 2017 08:47:54 +0000 (16:47 +0800)]
Merge pull request #2089 from alanbuxey/patch-4
update text (typo and grammar)
Arran Cudbard-Bell [Tue, 10 Oct 2017 03:28:16 +0000 (11:28 +0800)]
Merge pull request #2088 from jsynack/oct7updates
Misc updates
Arran Cudbard-Bell [Mon, 9 Oct 2017 15:58:38 +0000 (21:58 +0600)]
Don't close out the connection just signal it to reconnect if we can insert the FD
There are too many corner cases with freeing the connection within fd_active and fd_idle... They don't for example communicate to the caller that the connection is no longer valid.
Alan Buxey [Mon, 9 Oct 2017 14:03:02 +0000 (15:03 +0100)]
update text (typo and grammar)
Jason Ackley [Mon, 9 Oct 2017 13:46:46 +0000 (08:46 -0500)]
Misc Non-code impacting changes to docs
1. Start to sprinkle https vs. http to encourage a safer web.
Some of this had URLs redone to the wiki as the SSL redirect from freeradius.org
doesn't appear to be consistent. If this is desired then there are more.
2. Misc corrections to links that had become outdated or stale.
3. A few filename renames ( -> .md files)
4. Remove reference to google search box in the mailing list - couldnt find it?
5. Instead of talking about the age of the book - just give the publication date.
6. A few extra bash highlight-isms
Normally I would not batch so much into one commit - but these are all non-code impacting.
Jason Ackley [Mon, 9 Oct 2017 12:59:07 +0000 (07:59 -0500)]
Random adjustments and low-hanging fruit
Jason Ackley [Sat, 7 Oct 2017 16:43:40 +0000 (11:43 -0500)]
FreeADIUS -> FreeRADIUS :)
Jason Ackley [Sat, 7 Oct 2017 16:43:07 +0000 (11:43 -0500)]
Add Foundry-COA-Command (overlaps with existing)
Arran Cudbard-Bell [Thu, 5 Oct 2017 13:08:33 +0000 (21:08 +0800)]
Vendor not used
Arran Cudbard-Bell [Thu, 5 Oct 2017 12:35:27 +0000 (20:35 +0800)]
Move dict xlats into rlm_dict
Remove vendor specification from fr_dict_attr_by_oid as vendors are structural attributes now and it's not required
Arran Cudbard-Bell [Thu, 5 Oct 2017 08:03:18 +0000 (16:03 +0800)]
Revert "Fix a couple of log messages"
This reverts commit
eb5bce9fd8648e682ecd3cb761d152f2a90bad30 .
Arran Cudbard-Bell [Thu, 5 Oct 2017 02:12:00 +0000 (10:12 +0800)]
Fix a couple of log messages
Arran Cudbard-Bell [Wed, 4 Oct 2017 07:23:45 +0000 (14:23 +0700)]
Populate request packet timestamp
Arran Cudbard-Bell [Wed, 4 Oct 2017 02:24:26 +0000 (09:24 +0700)]
More cleanups in udp_request_free
Alan T. DeKok [Tue, 3 Oct 2017 20:10:03 +0000 (16:10 -0400)]
only call timer delete if we have a timer
Alan T. DeKok [Tue, 3 Oct 2017 20:09:48 +0000 (16:09 -0400)]
don't molest ev
Alan T. DeKok [Tue, 3 Oct 2017 18:50:51 +0000 (14:50 -0400)]
add time values to debug message
Alan T. DeKok [Tue, 3 Oct 2017 18:47:12 +0000 (14:47 -0400)]
set *ev->parent = NULL in the talloc destructor
so that the callers expectations are held true... that we can
just talloc_free() the pointer, and expect everything to be
cleaned up.
Alan T. DeKok [Tue, 3 Oct 2017 18:24:36 +0000 (14:24 -0400)]
print when retransmit timers hit, even if there's no connection
Alan T. DeKok [Tue, 3 Oct 2017 18:19:32 +0000 (14:19 -0400)]
assert that the parent is pointing to us
Alan T. DeKok [Tue, 3 Oct 2017 16:42:39 +0000 (12:42 -0400)]
move timer insert to udp function
Alan T. DeKok [Tue, 3 Oct 2017 16:26:31 +0000 (12:26 -0400)]
more debugging and "finished" flag
Alan T. DeKok [Tue, 3 Oct 2017 15:49:09 +0000 (11:49 -0400)]
enable RADIUS module by default
Arran Cudbard-Bell [Tue, 3 Oct 2017 15:53:48 +0000 (23:53 +0800)]
Restore previous state
Arran Cudbard-Bell [Tue, 3 Oct 2017 15:52:18 +0000 (23:52 +0800)]
Follow the normal state transition if we have a ->failed callback and it returns FR_CONNECTION_STATE_INIT
Alan T. DeKok [Tue, 3 Oct 2017 15:43:20 +0000 (11:43 -0400)]
remove it from the thread queue, too
Alan T. DeKok [Tue, 3 Oct 2017 15:37:11 +0000 (11:37 -0400)]
remove packet from c->queued on timeout or reply
Alan T. DeKok [Tue, 3 Oct 2017 15:29:08 +0000 (11:29 -0400)]
don't access u->rr if it doesn't exist
Alan T. DeKok [Tue, 3 Oct 2017 15:25:40 +0000 (11:25 -0400)]
print time after the timer has been initialized
Alan T. DeKok [Tue, 3 Oct 2017 15:24:13 +0000 (11:24 -0400)]
access u->timer directly
Alan T. DeKok [Tue, 3 Oct 2017 15:24:06 +0000 (11:24 -0400)]
notes for todo
Arran Cudbard-Bell [Tue, 3 Oct 2017 15:21:03 +0000 (23:21 +0800)]
Don't need this in _conn_failed as _conn_close is always called
Alan T. DeKok [Tue, 3 Oct 2017 15:12:10 +0000 (11:12 -0400)]
remove connection from list / heap
Alan T. DeKok [Tue, 3 Oct 2017 12:49:42 +0000 (08:49 -0400)]
don't use one-character variable names
or prefixes. or _p where it's not necessary
simplify some logic to make it clearer, and update comments
to describe what the code is doing