]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Mon, 25 Sep 2017 15:13:33 +0000 (11:13 -0400)]
move client src_ipaddr mangling to proto_radius
so that the process functions are entirely protocol agnostic
almost... they still print src/dst IP information for debugging
Alan T. DeKok [Mon, 25 Sep 2017 15:10:39 +0000 (11:10 -0400)]
start of new detail file listener
Alan T. DeKok [Mon, 25 Sep 2017 15:10:20 +0000 (11:10 -0400)]
moved old files out of the way
Alan T. DeKok [Mon, 25 Sep 2017 15:01:55 +0000 (11:01 -0400)]
minor cleanups and typos
Alan T. DeKok [Mon, 25 Sep 2017 14:39:45 +0000 (10:39 -0400)]
move detail.h to proto_detail
No one else needs it, or should use it.
Alan T. DeKok [Mon, 25 Sep 2017 14:38:17 +0000 (10:38 -0400)]
always write to the detail file
even if we read it from the detail file.
This change is necessary to encapsulate the new proto_detail code.
We can later add an attribute which passes the filename to
rlm_detail, so it can suppress infinite loops
Alan T. DeKok [Sun, 24 Sep 2017 20:15:52 +0000 (16:15 -0400)]
now that the code is smarter, we need less documentation
Alan T. DeKok [Sun, 24 Sep 2017 20:15:37 +0000 (16:15 -0400)]
use '=' instead of '==' for portability
Alan T. DeKok [Sun, 24 Sep 2017 20:14:00 +0000 (16:14 -0400)]
remove deprecated disable_tls* configuration
Alan T. DeKok [Sun, 24 Sep 2017 18:35:19 +0000 (14:35 -0400)]
use '=' instead of '==' for portability
Arran Cudbard-Bell [Sun, 24 Sep 2017 15:09:39 +0000 (23:09 +0800)]
More ifdefs
Arran Cudbard-Bell [Sun, 24 Sep 2017 14:56:24 +0000 (22:56 +0800)]
NOTE_REVOKE and NOTE_FUNLOCK not implemented by libkqueue
Arran Cudbard-Bell [Sun, 24 Sep 2017 14:25:41 +0000 (22:25 +0800)]
Support all vnode note types
Extends the event API to allow callbacks to be registered for all vnode NOTE_* events.
Alan T. DeKok [Sun, 24 Sep 2017 13:09:00 +0000 (09:09 -0400)]
manually disable TLSv1.2
Alan T. DeKok [Sun, 24 Sep 2017 13:00:28 +0000 (09:00 -0400)]
in v4 the 'type = fast' enables the module
not the config subsection. So we can always uncomment the
modules
Alan T. DeKok [Sun, 24 Sep 2017 13:00:00 +0000 (09:00 -0400)]
allow FAST to over-ride cipher_list
Alan T. DeKok [Sun, 24 Sep 2017 12:33:39 +0000 (08:33 -0400)]
root unknown from parent, not from dict root
Alan T. DeKok [Sat, 23 Sep 2017 17:03:07 +0000 (13:03 -0400)]
don't leak vp
Alan T. DeKok [Sat, 23 Sep 2017 13:20:57 +0000 (09:20 -0400)]
EAP-FAST-Crypto-Binding is 'octets'
until such time as eap_fast_decode_pair() decodes STRUCT types
Alan T. DeKok [Sat, 23 Sep 2017 13:18:19 +0000 (09:18 -0400)]
it helps to add the VPs to the cursor
Alan T. DeKok [Sat, 23 Sep 2017 13:01:40 +0000 (09:01 -0400)]
use decoded len, not len from all of the packet
Alan T. DeKok [Sat, 23 Sep 2017 12:46:47 +0000 (08:46 -0400)]
suppress bad error for EAP-FAST
instead of printing a message saying to ignore it
Alan T. DeKok [Sat, 23 Sep 2017 12:25:38 +0000 (08:25 -0400)]
tls_max_version cannot be less than 1.0
Isaac Boukris [Sat, 23 Sep 2017 11:01:10 +0000 (14:01 +0300)]
tls: some more casting to float
Signed-off-by: Isaac Boukris <iboukris@gmail.com>
Alan T. DeKok [Sat, 23 Sep 2017 01:54:57 +0000 (21:54 -0400)]
use v4 API
Isaac Boukris [Sat, 23 Sep 2017 00:01:32 +0000 (03:01 +0300)]
EAP-FAST: Fix EAP-FAST-MSCHAPv2 in anonymous tunnel (v3.0.x)
According to RFC 5422, in anonymous FAST tunnel the mschapv2
challenge is generated differently, from the phase1 exchange.
While the eap-fast module takes care to save internally the
peer-challenge to use, no one cares to read it before doing
the mschap check, causing it to fail.
Fix this by checking for PW_MS_CHAP_PEER_CHALLENGE and using it.
Signed-off-by: Isaac Boukris <iboukris@gmail.com>
Alan T. DeKok [Sat, 23 Sep 2017 01:46:49 +0000 (21:46 -0400)]
recurse on decode TLVs, instead of !TLV
and decode the contents of the TLV, instead of TLV itself
Alan T. DeKok [Sat, 23 Sep 2017 01:46:41 +0000 (21:46 -0400)]
cast to float
Alan T. DeKok [Fri, 22 Sep 2017 14:59:42 +0000 (10:59 -0400)]
note on todo
Alan T. DeKok [Fri, 22 Sep 2017 13:32:14 +0000 (09:32 -0400)]
track deleted sockets
and delay deleting them until all of the replies have been handled
Alan T. DeKok [Thu, 21 Sep 2017 20:01:26 +0000 (16:01 -0400)]
ensure we listen only on directories for vnodes
Alan T. DeKok [Thu, 21 Sep 2017 19:56:59 +0000 (15:56 -0400)]
add vnode to app_io, and notify it when things change
Alan T. DeKok [Thu, 21 Sep 2017 18:28:41 +0000 (14:28 -0400)]
doxygen
Alan T. DeKok [Thu, 21 Sep 2017 18:08:22 +0000 (14:08 -0400)]
allow adding "watch directory" to the network side
the function called from the event loop is empty and does
nothing for now. Later, it has to call a new function in
the app_io, and tell it "something changed in the directory".
That function then needs to open the new detail file, and
(somehow?) get those packets back to the fr_network_t, for
forwarding to a worker.
Alan T. DeKok [Thu, 21 Sep 2017 17:59:52 +0000 (13:59 -0400)]
add support for EVFILT_VNODE
for now, hard-coded to NOTE_EXTEND.
i.e. for directories, and for proto_detail
Alan T. DeKok [Thu, 21 Sep 2017 17:07:33 +0000 (13:07 -0400)]
sort packets by original request time, too
Alan T. DeKok [Thu, 21 Sep 2017 17:04:10 +0000 (13:04 -0400)]
keep the first request pending
so that if it's partially written, we won't over-write it with
a higher priority packet.
After that, replies are sent out ordered by priority, and
original recv time
Alan T. DeKok [Thu, 21 Sep 2017 16:58:01 +0000 (12:58 -0400)]
move waiting queue to heap
Alan T. DeKok [Thu, 21 Sep 2017 15:57:48 +0000 (11:57 -0400)]
note that we either write ALL of the data, or NONE of it
Alan T. DeKok [Thu, 21 Sep 2017 15:51:00 +0000 (11:51 -0400)]
if we get EWOULDBLOCK on the write, add packet to outbound queue
and enable write callback for the FD
Alan T. DeKok [Thu, 21 Sep 2017 15:30:14 +0000 (11:30 -0400)]
brackets are useful
Alan T. DeKok [Thu, 21 Sep 2017 15:26:58 +0000 (11:26 -0400)]
add a write queue to network sockets
Alan T. DeKok [Thu, 21 Sep 2017 15:12:42 +0000 (11:12 -0400)]
udp_send() may return an error
Alan T. DeKok [Tue, 19 Sep 2017 15:53:56 +0000 (11:53 -0400)]
remove unused enum
Alan T. DeKok [Tue, 19 Sep 2017 13:38:56 +0000 (09:38 -0400)]
remove unused variable
Alan T. DeKok [Tue, 19 Sep 2017 13:37:16 +0000 (09:37 -0400)]
update comments, and hoist check_config to earlier
so that we don't do work which is later ignored...
Arran Cudbard-Bell [Wed, 20 Sep 2017 03:07:47 +0000 (10:07 +0700)]
Don't create empty stripped-user-domain attributes
Arran Cudbard-Bell [Tue, 19 Sep 2017 11:06:24 +0000 (18:06 +0700)]
Merge pull request #2075 from herwinw/rlm_perl_cleanup
Small cleanup in rlm_perl
Herwin Weststrate [Tue, 19 Sep 2017 11:04:02 +0000 (13:04 +0200)]
Small cleanup in rlm_perl
Arran Cudbard-Bell [Tue, 19 Sep 2017 05:51:41 +0000 (12:51 +0700)]
Add ubuntu16, fix libnl install calls
Alan T. DeKok [Mon, 18 Sep 2017 20:45:02 +0000 (16:45 -0400)]
open local sockets before initializing modules. Fixes #1691
Herwin Weststrate [Mon, 18 Sep 2017 06:52:42 +0000 (08:52 +0200)]
More graceful handling in rlm_perl on die in called method
This should fix #2067
Alan T. DeKok [Sun, 17 Sep 2017 12:45:14 +0000 (08:45 -0400)]
remove "stack" as a separate argument
it's always request->stack.
The original idea was to allow different interpreters for the
same request. But practice shows that it's much better to just
have different requests instead.
Alan T. DeKok [Sun, 17 Sep 2017 12:37:27 +0000 (08:37 -0400)]
notes on todo's
Alan DeKok [Fri, 15 Sep 2017 19:31:19 +0000 (15:31 -0400)]
Merge pull request #2071 from jimdigriz/v4.0.x-rhspec
V4.0.x rhspec
Alan T. DeKok [Fri, 15 Sep 2017 17:32:55 +0000 (13:32 -0400)]
enforce max request lifetime on the child request
Alan T. DeKok [Fri, 15 Sep 2017 17:31:50 +0000 (13:31 -0400)]
correct number ranges in comments
Alexander Clouter [Fri, 15 Sep 2017 14:11:57 +0000 (15:11 +0100)]
update rhspec
Alexander Clouter [Fri, 15 Sep 2017 14:11:35 +0000 (15:11 +0100)]
centos6 docker build fix
Alan T. DeKok [Fri, 15 Sep 2017 13:32:50 +0000 (09:32 -0400)]
shrink unlang_group_t slightly
with anonymous struct anonymous union
Matthew Newton [Fri, 15 Sep 2017 13:01:12 +0000 (14:01 +0100)]
people try to use client certificates, and it generally doesn't work
Arran Cudbard-Bell [Fri, 15 Sep 2017 08:22:41 +0000 (15:22 +0700)]
Add iterator for iterating by DA
Arran Cudbard-Bell [Fri, 15 Sep 2017 07:25:41 +0000 (14:25 +0700)]
Remove the VP from the cursor before freeing
That way the cursor doesn't contain references to already freed memory.
Arran Cudbard-Bell [Fri, 15 Sep 2017 06:05:04 +0000 (13:05 +0700)]
Swap out more calls to use the new cursor api
Arran Cudbard-Bell [Fri, 15 Sep 2017 01:36:27 +0000 (08:36 +0700)]
Use correct signed integer field in rlm_lua
Arran Cudbard-Bell [Fri, 15 Sep 2017 01:34:28 +0000 (08:34 +0700)]
Replace more vp_cursor_t calls with fr_cursor_t calls
Alan T. DeKok [Thu, 14 Sep 2017 23:55:11 +0000 (19:55 -0400)]
no unused variable for you
Alan T. DeKok [Thu, 14 Sep 2017 23:29:58 +0000 (19:29 -0400)]
hack together "call"
it doesn't do yeild, and it overloads things it shouldn't overload
but it works. It can be cleaned up later
Alan T. DeKok [Thu, 14 Sep 2017 23:26:42 +0000 (19:26 -0400)]
don't print out TLS tunnel stuff... it's not relevant
We need to find a better way to do this. Perhaps print out
the parent / child virtual server?
Alan T. DeKok [Thu, 14 Sep 2017 23:09:18 +0000 (19:09 -0400)]
use correct process ptr
Alan T. DeKok [Thu, 14 Sep 2017 17:49:05 +0000 (13:49 -0400)]
detach the child before freeing it
Alan T. DeKok [Thu, 14 Sep 2017 17:47:00 +0000 (13:47 -0400)]
"call" can only be used in a "subrequest" section.
For now, until we get more code written
Alan T. DeKok [Thu, 14 Sep 2017 17:22:51 +0000 (13:22 -0400)]
verify that server.PACKET-Type exists
Alan T. DeKok [Thu, 14 Sep 2017 17:21:49 +0000 (13:21 -0400)]
moved "cache process ptr" to mod_bootstrap
so it's available for later checks
Alan T. DeKok [Thu, 14 Sep 2017 16:53:23 +0000 (12:53 -0400)]
add UNLANG_TYPE_CALL to the parser
and bare-bones interpreter
Alan T. DeKok [Thu, 14 Sep 2017 15:55:07 +0000 (11:55 -0400)]
it's compile_subrequest(), not compile_create()
Alan T. DeKok [Thu, 14 Sep 2017 15:53:44 +0000 (11:53 -0400)]
Cache process function by packet name
Alan T. DeKok [Thu, 14 Sep 2017 14:48:46 +0000 (10:48 -0400)]
don't log Cleartext-Password
Alan T. DeKok [Thu, 14 Sep 2017 14:48:01 +0000 (10:48 -0400)]
document subrequest and detach
Arran Cudbard-Bell [Thu, 14 Sep 2017 16:33:05 +0000 (23:33 +0700)]
Happily not needed
...as the new cursor API was designed somewhat more than the original, and doesn't require dumb hacks.
Arran Cudbard-Bell [Thu, 14 Sep 2017 14:33:16 +0000 (21:33 +0700)]
Sprinkle more verify calls
Arran Cudbard-Bell [Thu, 14 Sep 2017 14:31:18 +0000 (21:31 +0700)]
Formatting
Arran Cudbard-Bell [Thu, 14 Sep 2017 12:42:00 +0000 (19:42 +0700)]
Switch to the new cursor API for tmpl_cursors
This means we can remove tmpl_cursor_next, and pass around cursors which represents subsets of attribute lists (without copying VALUE_PAIRs), which is nice...
Arran Cudbard-Bell [Thu, 14 Sep 2017 12:22:40 +0000 (19:22 +0700)]
Typo in cursor_replace
Arran Cudbard-Bell [Thu, 14 Sep 2017 08:49:08 +0000 (15:49 +0700)]
Pick up brewed ssl for eapol_test
Arran Cudbard-Bell [Thu, 14 Sep 2017 11:26:59 +0000 (18:26 +0700)]
Merge pull request #2065 from pwdng/v4.0.x
Fix RPM spec file
Philippe Wooding [Thu, 14 Sep 2017 11:06:01 +0000 (13:06 +0200)]
Fix RPM spec file
Alan T. DeKok [Wed, 13 Sep 2017 19:46:13 +0000 (15:46 -0400)]
hack for NUM_ALL
Not sure why the xlat-list test *ever* passed...
Alan T. DeKok [Wed, 13 Sep 2017 19:25:25 +0000 (15:25 -0400)]
more test_fail
Alan T. DeKok [Wed, 13 Sep 2017 18:52:45 +0000 (14:52 -0400)]
use "success" instead of manual thing
Alan T. DeKok [Wed, 13 Sep 2017 18:49:54 +0000 (14:49 -0400)]
move to using test_fail macro
Multi-line perl replace is your friend...
Alan T. DeKok [Wed, 13 Sep 2017 18:33:58 +0000 (14:33 -0400)]
add test_fail
which shows the line number of the block which called the
test_fail policy
Alan T. DeKok [Wed, 13 Sep 2017 18:31:02 +0000 (14:31 -0400)]
added %{interpreter:...} xlat
for introspection of the interpreter state
Arran Cudbard-Bell [Wed, 13 Sep 2017 18:32:20 +0000 (01:32 +0700)]
Arran Cudbard-Bell [Wed, 13 Sep 2017 18:15:43 +0000 (01:15 +0700)]
See if stages work any better now
Alan T. DeKok [Wed, 13 Sep 2017 18:12:11 +0000 (14:12 -0400)]
clean up compile_action_pair
and require action over-rides to be an "actions" section
Alan T. DeKok [Wed, 13 Sep 2017 17:47:01 +0000 (13:47 -0400)]
'actions' must be the LAST thing in a group
Alan T. DeKok [Wed, 13 Sep 2017 17:46:44 +0000 (13:46 -0400)]
'actions' may be in a "group"
Alan T. DeKok [Wed, 13 Sep 2017 16:59:55 +0000 (12:59 -0400)]
disallow action over-rides in random places
Alan T. DeKok [Wed, 13 Sep 2017 16:38:00 +0000 (12:38 -0400)]
don't install rlm_test or rlm_example