]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Tue, 21 Nov 2017 16:45:09 +0000 (11:45 -0500)]
start of dynamic_client_alloc()
Alan T. DeKok [Tue, 21 Nov 2017 16:27:43 +0000 (11:27 -0500)]
start of dynamic client code
Arran Cudbard-Bell [Tue, 21 Nov 2017 16:29:17 +0000 (16:29 +0000)]
...because you know static analysis is going to flag that at some point *sigh*
Arran Cudbard-Bell [Tue, 21 Nov 2017 16:26:18 +0000 (16:26 +0000)]
Use current result not frame->result when yielding
and don't mess with frame->result
Arran Cudbard-Bell [Tue, 21 Nov 2017 16:22:00 +0000 (16:22 +0000)]
Revert "Set frame->result when yielding"
This reverts commit
1f7ceeb44c90809c668d68f07e749c85c52a9f8b .
Alan T. DeKok [Tue, 21 Nov 2017 16:19:19 +0000 (11:19 -0500)]
there is always a module return code.
This goes along with commit
a262092bf41c
Alan T. DeKok [Tue, 21 Nov 2017 16:12:24 +0000 (11:12 -0500)]
fix for EAP-FAST.
The parent TLV has been decoded into a "struct". So we can't
check the sizes. And the sizes of the various fields are 8-bit,
not 32-bit
Alan T. DeKok [Mon, 20 Nov 2017 20:09:08 +0000 (15:09 -0500)]
fmt is parented from node, not node->fmt
Alan T. DeKok [Tue, 21 Nov 2017 16:05:40 +0000 (11:05 -0500)]
allow COMBO_IP_PREFIX too
Alan T. DeKok [Tue, 21 Nov 2017 15:49:49 +0000 (10:49 -0500)]
request->rcode is return code from last module call
Alan T. DeKok [Tue, 21 Nov 2017 15:45:44 +0000 (10:45 -0500)]
0 is RLM_MODULE_REJECT, which is a valid rcode
Alan T. DeKok [Mon, 20 Nov 2017 21:00:03 +0000 (16:00 -0500)]
add combo IP address / prefix
Arran Cudbard-Bell [Tue, 21 Nov 2017 15:08:20 +0000 (15:08 +0000)]
clang scan fix
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:50:29 +0000 (14:50 +0000)]
Add a test for yielding within a parallel section
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:50:05 +0000 (14:50 +0000)]
Add a delay module that we can use to force requests to yield and be immediately rescheduled
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:49:36 +0000 (14:49 +0000)]
Check the current rcode not the frame rcode for the UNLANG_ACTION_YIELD hack
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:49:04 +0000 (14:49 +0000)]
Check for an attempt to pop the top frame and deal with that explicitly by breaking out the interpreter loop
This allows the function to return at the right point without decrementing stack->depth
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:48:06 +0000 (14:48 +0000)]
Formatting
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:47:07 +0000 (14:47 +0000)]
Flesh out unlang_interpret_synchronous some more so it has a fake backlog to go with its fake event list
Arran Cudbard-Bell [Tue, 21 Nov 2017 14:45:19 +0000 (14:45 +0000)]
Better name
Arran Cudbard-Bell [Tue, 21 Nov 2017 12:21:13 +0000 (12:21 +0000)]
Set frame->result when yielding
Arran Cudbard-Bell [Tue, 21 Nov 2017 11:50:07 +0000 (11:50 +0000)]
Ignore bad binding attributes
Arran Cudbard-Bell [Tue, 21 Nov 2017 11:03:31 +0000 (11:03 +0000)]
Use a proper event loop in unit_test_module.c and remove proto_radius from the test path
We're testing unlang keywords, not the proto_radius module
Arran Cudbard-Bell [Tue, 21 Nov 2017 10:46:44 +0000 (10:46 +0000)]
Allow zero delay in rlm_delay (just forces the request to be rescheduled)
Alan T. DeKok [Mon, 20 Nov 2017 18:53:28 +0000 (13:53 -0500)]
set THREADID callbacks instead of global callbacks
when we have them.
Alan T. DeKok [Fri, 17 Nov 2017 21:05:33 +0000 (16:05 -0500)]
eap-fast: fail on crypto-binding mismatch
as per rfc 4851, an invalid crypto-binding tlv is a
fatal error, handle accordingly.
Signed-off-by: Isaac Boukris <iboukris@gmail.com>
Alan T. DeKok [Mon, 20 Nov 2017 18:42:58 +0000 (13:42 -0500)]
check sizes, and port fixes from 3.0
Matthew Newton [Mon, 20 Nov 2017 18:41:00 +0000 (18:41 +0000)]
Update dockerbuild script to match Arran's new names
Arran Cudbard-Bell [Mon, 20 Nov 2017 16:05:06 +0000 (16:05 +0000)]
Rename the Dockerfile that builds the dependencies to Dockerfile.deps not Dockerfile.source
Alan T. DeKok [Sat, 18 Nov 2017 14:29:08 +0000 (09:29 -0500)]
don't add quotes. GNU Make is very literal about it
and doesn't need them
Arran Cudbard-Bell [Mon, 20 Nov 2017 15:48:15 +0000 (15:48 +0000)]
Pull libkqueue from the official freeradius fork
Arran Cudbard-Bell [Mon, 20 Nov 2017 15:05:16 +0000 (15:05 +0000)]
Arran Cudbard-Bell [Sun, 19 Nov 2017 11:32:23 +0000 (11:32 +0000)]
Add registration functions and a new prototype for async xlats
Arran Cudbard-Bell [Sun, 19 Nov 2017 10:56:49 +0000 (10:56 +0000)]
Fix strict-prototypes warnings in MySQL
Arran Cudbard-Bell [Sat, 18 Nov 2017 20:21:11 +0000 (20:21 +0000)]
Add xlat_init function
Arran Cudbard-Bell [Sat, 18 Nov 2017 13:03:23 +0000 (13:03 +0000)]
Define mem macro in control_test.c
Arran Cudbard-Bell [Sat, 18 Nov 2017 13:03:13 +0000 (13:03 +0000)]
Formatting
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:39:38 +0000 (23:39 +0000)]
Move MEM to debug.h so we can use it in utilities
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:39:22 +0000 (23:39 +0000)]
Use EXIT_FAILURE for portability
Apparently not all systems use 1 to indicate failure. C99 says "Finally, control is returned to the host environment. If the value of status is zero or EXIT_SUCCESS, an implementation-defined form of the status successful termination is returned. If the value of status is EXIT_FAILURE, an implementation-defined form of the status unsuccessful termination is returned. Otherwise the status returned is implementation-defined."
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:31:47 +0000 (23:31 +0000)]
Memory allocation check
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:31:30 +0000 (23:31 +0000)]
Memory allocation check, and call fr_exit_now not _exit
So we know where we exited...
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:29:38 +0000 (23:29 +0000)]
Fix inappropriate use of exit in proto_bfd
Arran Cudbard-Bell [Fri, 17 Nov 2017 23:29:18 +0000 (23:29 +0000)]
Use mem macros for memory allocation errors
Alan T. DeKok [Fri, 17 Nov 2017 22:56:24 +0000 (17:56 -0500)]
memory allocation. CID #
1420942
Alan T. DeKok [Fri, 17 Nov 2017 22:55:38 +0000 (17:55 -0500)]
memory allocation. CID #
1420935
Alan T. DeKok [Fri, 17 Nov 2017 22:54:39 +0000 (17:54 -0500)]
Alan T. DeKok [Fri, 17 Nov 2017 22:53:51 +0000 (17:53 -0500)]
memory allocation. CID #
1421314
Alan T. DeKok [Fri, 17 Nov 2017 22:52:48 +0000 (17:52 -0500)]
print errors and work filename
Alan T. DeKok [Fri, 17 Nov 2017 22:35:11 +0000 (17:35 -0500)]
ignore empty detail files
Alan T. DeKok [Fri, 17 Nov 2017 20:29:18 +0000 (15:29 -0500)]
differentiate ENOENT from other errors
Alan T. DeKok [Fri, 17 Nov 2017 20:19:00 +0000 (15:19 -0500)]
mod_close() removes FDs from the event loop when it closes
network.c closes the IO handle. But the VNODE handle is added
by proto_detail_work
Alan T. DeKok [Fri, 17 Nov 2017 20:09:36 +0000 (15:09 -0500)]
clean up error path
Arran Cudbard-Bell [Fri, 17 Nov 2017 18:32:02 +0000 (18:32 +0000)]
Fix all those 32s that should be 16s
Arran Cudbard-Bell [Fri, 17 Nov 2017 18:08:31 +0000 (18:08 +0000)]
Fix the dictionary types
Arran Cudbard-Bell [Fri, 17 Nov 2017 18:08:18 +0000 (18:08 +0000)]
Typo
Arran Cudbard-Bell [Fri, 17 Nov 2017 17:56:52 +0000 (17:56 +0000)]
The power of C11 compells you
Arran Cudbard-Bell [Fri, 17 Nov 2017 17:41:14 +0000 (17:41 +0000)]
Simplify code in detail_write
Could compress it further by selecting src/dst da, but not sure if AF is ever 'null' i.e. no ip address.
Alan T. DeKok [Fri, 17 Nov 2017 17:36:23 +0000 (12:36 -0500)]
can we be done now? Please?
Alan T. DeKok [Fri, 17 Nov 2017 17:31:52 +0000 (12:31 -0500)]
should really move this nonsense to value boxes
Alan T. DeKok [Fri, 17 Nov 2017 17:29:02 +0000 (12:29 -0500)]
set AF too
Alan T. DeKok [Fri, 17 Nov 2017 17:16:38 +0000 (12:16 -0500)]
don't include trailing 0
Alan T. DeKok [Fri, 17 Nov 2017 16:23:04 +0000 (11:23 -0500)]
check Message-Authenticator correctly
Alan T. DeKok [Fri, 17 Nov 2017 15:11:08 +0000 (10:11 -0500)]
don't crash on <internal>
Alan T. DeKok [Fri, 17 Nov 2017 15:08:13 +0000 (10:08 -0500)]
set data.type, too
Alan T. DeKok [Thu, 16 Nov 2017 18:36:33 +0000 (13:36 -0500)]
reorder
Arran Cudbard-Bell [Thu, 16 Nov 2017 22:46:05 +0000 (22:46 +0000)]
Validate requests too in radsniff
Arran Cudbard-Bell [Thu, 16 Nov 2017 22:45:50 +0000 (22:45 +0000)]
Return an error if we can't validate the request authenticator (oops)
Alan T. DeKok [Thu, 16 Nov 2017 17:10:20 +0000 (12:10 -0500)]
decrement workers when we exit
Alan T. DeKok [Thu, 16 Nov 2017 17:08:56 +0000 (12:08 -0500)]
call worker close, and don't just close the FD
Alan T. DeKok [Thu, 16 Nov 2017 16:38:17 +0000 (11:38 -0500)]
touch before free, not after
Alan T. DeKok [Thu, 16 Nov 2017 16:08:29 +0000 (11:08 -0500)]
typo
Alan T. DeKok [Thu, 16 Nov 2017 15:53:57 +0000 (10:53 -0500)]
have the worker clean up it's IO instance, not the file module
and in work_init(), poll for changes if the worker is still alive.
Alan T. DeKok [Thu, 16 Nov 2017 01:53:19 +0000 (20:53 -0500)]
travis fixes
Alan T. DeKok [Thu, 16 Nov 2017 00:04:02 +0000 (19:04 -0500)]
call fr_network_listen_read() on Linux
because libkqueue is broken
Alan T. DeKok [Wed, 15 Nov 2017 23:55:10 +0000 (18:55 -0500)]
add 'nr' to event_list_set()
so that if a reader needs to poke the network, it can do so
Alan T. DeKok [Wed, 15 Nov 2017 21:55:04 +0000 (16:55 -0500)]
don't do work_init() twice
Alan T. DeKok [Wed, 15 Nov 2017 21:37:45 +0000 (16:37 -0500)]
fixes for building on linux
Alan T. DeKok [Wed, 15 Nov 2017 18:31:37 +0000 (13:31 -0500)]
clean up code and remove goto
Alan T. DeKok [Wed, 15 Nov 2017 18:24:08 +0000 (13:24 -0500)]
try renaming detail* to detail.work only once
Alan T. DeKok [Wed, 15 Nov 2017 18:20:31 +0000 (13:20 -0500)]
allow poll_interval == 0 on Linux
because inotify works there <sigh>
Alan T. DeKok [Wed, 15 Nov 2017 18:18:12 +0000 (13:18 -0500)]
if no worker and the dir changed, start checking again
Alan T. DeKok [Wed, 15 Nov 2017 18:07:11 +0000 (13:07 -0500)]
track the number of workers via a mutex-protected counter
Alan T. DeKok [Wed, 15 Nov 2017 16:05:30 +0000 (11:05 -0500)]
delay polling until such time as /proc/PID changes permission
Alan T. DeKok [Wed, 15 Nov 2017 14:45:59 +0000 (09:45 -0500)]
Ensure that libkqueue / Linux doesn't bork us.
The kqueue API takes an FD, but inotify requires a filename.
libkqueue uses /proc/PID/fd/# to look up the FD -> filename mapping.
However, if you start the server as "root", and then swap to "radiusd",
/proc/PID will be owned by "root" for security reasons. The only way
to make /proc/PID owned by "radiusd" is to set the DUMPABLE flag.
Instead of making the poor sysadmin figure this out,
we check for this situation, and give them a
descriptive message telling them what to do.
Alan T. DeKok [Wed, 15 Nov 2017 14:40:21 +0000 (09:40 -0500)]
expose uid / gid / core dumps, etc.
Alan T. DeKok [Wed, 15 Nov 2017 14:30:49 +0000 (09:30 -0500)]
0 is also allowed
Arran Cudbard-Bell [Wed, 15 Nov 2017 09:16:02 +0000 (09:16 +0000)]
Regenerate rlm_rest configure script
Arran Cudbard-Bell [Tue, 14 Nov 2017 23:14:55 +0000 (23:14 +0000)]
Doxygen
Arran Cudbard-Bell [Tue, 14 Nov 2017 23:12:38 +0000 (23:12 +0000)]
Fix stupid non-issue flagged by GCC
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:54:58 +0000 (22:54 +0000)]
First stage of async xlats
Todo:
- Add custom async xlat function resumption type
- Test pushing xlats from modules
- Add new type of async xlat callback to expose additional functionality
- Do something sane with intermediary contexts for value boxes
- Convert tmpl expansion to also be async
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:52:20 +0000 (22:52 +0000)]
don't dup the closing brace in xlat_tokenize_expansion
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:51:24 +0000 (22:51 +0000)]
More xlat inline tests
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:51:13 +0000 (22:51 +0000)]
Shuffle debug statements around between functions in xlat_eval.c
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:48:00 +0000 (22:48 +0000)]
Fix all asprintf calls so they set the correct type
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:45:40 +0000 (22:45 +0000)]
Docs
Arran Cudbard-Bell [Tue, 14 Nov 2017 22:45:29 +0000 (22:45 +0000)]
Set the correct type forregex subcaptures
Arran Cudbard-Bell [Tue, 14 Nov 2017 21:59:26 +0000 (21:59 +0000)]
Use custom realloc function for strings
Ensures the type is correct and \0 terminates
Arran Cudbard-Bell [Tue, 14 Nov 2017 21:58:54 +0000 (21:58 +0000)]
Use typed_asprintf in sql_xlat
Alan T. DeKok [Tue, 14 Nov 2017 20:18:42 +0000 (15:18 -0500)]
just assign DHCP-Your-IP-Address natively
Alan T. DeKok [Tue, 14 Nov 2017 20:16:43 +0000 (15:16 -0500)]
make IP address assignment generic