]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Fri, 2 Nov 2018 13:43:21 +0000 (09:43 -0400)]
stop radmin only if it was started
Arran Cudbard-Bell [Fri, 2 Nov 2018 13:37:15 +0000 (09:37 -0400)]
Only do cap checks if we're not root
Create an error stack so that if the bind *does* fail, then there are helpful hints about what to do.
Alan T. DeKok [Fri, 2 Nov 2018 13:33:03 +0000 (09:33 -0400)]
fix for map_afrom_cs
Alan T. DeKok [Fri, 2 Nov 2018 12:26:36 +0000 (08:26 -0400)]
fix my email address from ox.org to freeradius.org
Alan T. DeKok [Fri, 2 Nov 2018 12:17:19 +0000 (08:17 -0400)]
add suid up/down for DHCP
Alan T. DeKok [Thu, 1 Nov 2018 19:05:35 +0000 (15:05 -0400)]
make map_afrom_cs() take a TALLOC_CTX argument
so that the maps aren't always allocated from the CONF_SECTION
Which is useful in some cases, and means that the API is similar
to other functions. e.g. map_afrom_cp()
Alan T. DeKok [Thu, 1 Nov 2018 18:41:35 +0000 (14:41 -0400)]
don't allow rlm_always to change it's running config
it was only used for one module test. We can now do the same
thing directly in unlang.
Arran Cudbard-Bell [Fri, 2 Nov 2018 01:27:58 +0000 (21:27 -0400)]
Check if we have the CAP_NET_BIND_SERVICE capability
If we don't, error out, and say how to set it
Arran Cudbard-Bell [Fri, 2 Nov 2018 01:27:12 +0000 (21:27 -0400)]
Rename some variables in fr_get_debug_state
Arran Cudbard-Bell [Thu, 1 Nov 2018 19:04:05 +0000 (15:04 -0400)]
Fixup argument names to be consistent between xlat functions
Arran Cudbard-Bell [Thu, 1 Nov 2018 17:48:28 +0000 (13:48 -0400)]
Unused but needed for dictionary work
Arran Cudbard-Bell [Thu, 1 Nov 2018 17:47:13 +0000 (13:47 -0400)]
Rename modules.[ch] to module.[ch]
Alan T. DeKok [Thu, 1 Nov 2018 16:39:56 +0000 (12:39 -0400)]
re-add module_thread_instance_by_data()
Alan T. DeKok [Thu, 1 Nov 2018 15:50:54 +0000 (11:50 -0400)]
remove "ctx" from fr_pair_list_move()
all of the callers already allocated the VPs from the correct
ctx, so there's no need to call fr_pair_steal(), which means
there's no need for the ctx
Alan T. DeKok [Thu, 1 Nov 2018 15:29:45 +0000 (11:29 -0400)]
remove unused functions
Alan T. DeKok [Thu, 1 Nov 2018 15:22:31 +0000 (11:22 -0400)]
use the correct parent, instead of calling steal()
Alan T. DeKok [Thu, 1 Nov 2018 15:02:36 +0000 (11:02 -0400)]
there's no need to talloc_steal() here.
The entry is parented from the instance anyways, so there's no
need to reparent the entry to be from the tree. The lifetime of
all of those objects is the same anyways: the lifetime of the
instance data.
Alan T. DeKok [Thu, 1 Nov 2018 14:41:06 +0000 (10:41 -0400)]
use an array in each thread for module instances
which should be faster and simpler.
Alan T. DeKok [Thu, 1 Nov 2018 14:10:14 +0000 (10:10 -0400)]
rename
because everything else is "foo.c" and "foo.h"
Alan T. DeKok [Thu, 1 Nov 2018 14:03:25 +0000 (10:03 -0400)]
allocate dl_loader from a new context
because we reference count proto_foo_* as new connections are
opened.
Alan T. DeKok [Thu, 1 Nov 2018 13:06:58 +0000 (09:06 -0400)]
use correct parent for talloc
Alan T. DeKok [Tue, 30 Oct 2018 19:24:47 +0000 (15:24 -0400)]
force all users of master IO to be thread-safe
Alan T. DeKok [Tue, 30 Oct 2018 19:20:12 +0000 (15:20 -0400)]
use 2M of mprotected memory for travis builds
Alan T. DeKok [Tue, 30 Oct 2018 15:43:52 +0000 (11:43 -0400)]
use NULL as talloc parent for thread_instance
Alan T. DeKok [Tue, 30 Oct 2018 15:28:28 +0000 (11:28 -0400)]
don't reset work_io_instance. It's now "const"
Alan T. DeKok [Tue, 30 Oct 2018 15:23:31 +0000 (11:23 -0400)]
move mutex into thread instance of the file reader
Alan T. DeKok [Tue, 30 Oct 2018 15:14:38 +0000 (11:14 -0400)]
notes on listener being thread-safe here
Alan T. DeKok [Tue, 30 Oct 2018 15:08:56 +0000 (11:08 -0400)]
move prototype
Alan T. DeKok [Tue, 30 Oct 2018 15:07:21 +0000 (11:07 -0400)]
allocate thread_instance from NULL ctx
and add a destructor
Alan T. DeKok [Tue, 30 Oct 2018 15:05:00 +0000 (11:05 -0400)]
rename "live" to "thread" for consistency
Arran Cudbard-Bell [Tue, 30 Oct 2018 13:52:07 +0000 (09:52 -0400)]
Better error
Arran Cudbard-Bell [Mon, 29 Oct 2018 23:05:28 +0000 (19:05 -0400)]
Remove spurious warning
Arran Cudbard-Bell [Mon, 29 Oct 2018 22:59:43 +0000 (18:59 -0400)]
lldb doesn't like this *sigh*
Arran Cudbard-Bell [Mon, 29 Oct 2018 22:52:25 +0000 (18:52 -0400)]
Tweak error
Arran Cudbard-Bell [Mon, 29 Oct 2018 22:20:54 +0000 (18:20 -0400)]
Terminate category
Alan T. DeKok [Mon, 29 Oct 2018 20:07:32 +0000 (16:07 -0400)]
this variable exists only in some builds
Alan T. DeKok [Mon, 29 Oct 2018 19:24:26 +0000 (15:24 -0400)]
no one uses this
so we delete it. There's no tests, no example module config,
no documentation, etc.
Alan T. DeKok [Mon, 29 Oct 2018 19:22:37 +0000 (15:22 -0400)]
remove last vestiges of struct signing
Alan T. DeKok [Mon, 29 Oct 2018 19:20:52 +0000 (15:20 -0400)]
we now always build with threads
Alan T. DeKok [Mon, 29 Oct 2018 19:20:26 +0000 (15:20 -0400)]
remove WITH_TCP flag.
We always build with TCP
Arran Cudbard-Bell [Mon, 29 Oct 2018 17:08:54 +0000 (13:08 -0400)]
Fix invalid/spurious leaks emitted from LSAN when test processes are forked
Alan T. DeKok [Sun, 28 Oct 2018 17:41:25 +0000 (13:41 -0400)]
always set src/dst ip/port
manual port of commit
1d090986ad0
Alan T. DeKok [Fri, 26 Oct 2018 17:12:52 +0000 (13:12 -0400)]
travis fixes
Alan T. DeKok [Fri, 26 Oct 2018 16:37:40 +0000 (12:37 -0400)]
convert detail reader to thread-safe.
And fix bug where when mod_write() unpauses the socket, it also
seeks to the start of the file. This ensures that if there is
leftover data in the buffer, mod_read() will be called again.
We can't rely on network.c to check "leftover", because it has
no idea if the socket "no data ready" means we should still call
mod_read(). For detail, yes. For other protocols, maybe no.
Also fix the work instantiation so that it's done once, and not
multiple times.
Alan T. DeKok [Thu, 25 Oct 2018 19:01:18 +0000 (15:01 -0400)]
make proto_control_unix thread-safe
Alan T. DeKok [Thu, 25 Oct 2018 18:52:57 +0000 (14:52 -0400)]
we need "const", too
Alan T. DeKok [Thu, 25 Oct 2018 18:46:34 +0000 (14:46 -0400)]
remove proto_control_tcp
proto_control_unix has many more features and works.
proto_control_tcp is missing many things.
Once proto_control_unix is finished, we can look at re-enabling
proto_control_tcp
Alan T. DeKok [Thu, 25 Oct 2018 18:45:23 +0000 (14:45 -0400)]
make proto_vmps_udp thread-safe
Alan T. DeKok [Thu, 25 Oct 2018 18:40:26 +0000 (14:40 -0400)]
cleanups
Alan T. DeKok [Thu, 25 Oct 2018 18:39:55 +0000 (14:39 -0400)]
cleanups, and convert TCP to thread-safe
Arran Cudbard-Bell [Wed, 24 Oct 2018 22:39:18 +0000 (18:39 -0400)]
Error out if parse callbacks fail or we can't push rules
Arran Cudbard-Bell [Fri, 19 Oct 2018 01:32:16 +0000 (21:32 -0400)]
Packet-Type is in ther internal dictionary
Arran Cudbard-Bell [Fri, 19 Oct 2018 01:41:47 +0000 (21:41 -0400)]
Fix typo in error
Arran Cudbard-Bell [Fri, 19 Oct 2018 14:39:21 +0000 (10:39 -0400)]
TLVs can be encrypted
Arran Cudbard-Bell [Fri, 19 Oct 2018 14:46:31 +0000 (10:46 -0400)]
Fixup tmpl validation
This checks that DAs are rooted correctly, and various other weird edge cases.
Arran Cudbard-Bell [Wed, 17 Oct 2018 19:12:53 +0000 (15:12 -0400)]
Missed free-on-error
Arran Cudbard-Bell [Wed, 24 Oct 2018 17:20:07 +0000 (13:20 -0400)]
Provide a case insensitive hash function
Arran Cudbard-Bell [Wed, 24 Oct 2018 17:25:06 +0000 (13:25 -0400)]
Mark VALUE_PAIR *req as unused in winbind_group_cmp
Arran Cudbard-Bell [Wed, 24 Oct 2018 17:22:41 +0000 (13:22 -0400)]
last_attr must be dictionary specific
Alan T. DeKok [Wed, 24 Oct 2018 17:07:46 +0000 (13:07 -0400)]
fix const / threading issues
Alan T. DeKok [Wed, 24 Oct 2018 16:41:13 +0000 (12:41 -0400)]
use thread-specific instance data
Arran Cudbard-Bell [Wed, 24 Oct 2018 16:21:34 +0000 (12:21 -0400)]
Add more types in by default
Alan T. DeKok [Wed, 24 Oct 2018 16:11:07 +0000 (12:11 -0400)]
use thread-specific instance data
Alan T. DeKok [Wed, 24 Oct 2018 15:26:30 +0000 (11:26 -0400)]
add thread_inst_size for app_io, and allocations for it.
Nothing uses that field, so this commit is safe.
The next set of commits will split out the thread/socket data
from the constant configuration data.
Alan T. DeKok [Wed, 24 Oct 2018 15:24:04 +0000 (11:24 -0400)]
comments on where / how sockets are opened
so that the documentation points to the correct location of the
source
Arran Cudbard-Bell [Mon, 22 Oct 2018 20:48:03 +0000 (16:48 -0400)]
Merge pull request #2336 from KI4STU/patch-1
Update filter: reject double dots
Detrick [Mon, 22 Oct 2018 18:33:58 +0000 (14:33 -0400)]
Update filter: reject double dots
Edit Module-Failure-Message to add clarity. Previous message ended with ".." which looks somewhat like an ellipsis giving the impression that the message has been truncated. New message replaces ".." with the word "dots" and adds an example.
Alan T. DeKok [Thu, 18 Oct 2018 16:21:10 +0000 (12:21 -0400)]
more remove HAVE_PTHREAD_H
Alan T. DeKok [Thu, 18 Oct 2018 16:19:25 +0000 (12:19 -0400)]
remove HAVE_PTHREAD_H
Alan T. DeKok [Thu, 18 Oct 2018 16:17:58 +0000 (12:17 -0400)]
more remove HAVE_PTHREAD_H
Alan T. DeKok [Thu, 18 Oct 2018 14:03:41 +0000 (10:03 -0400)]
rely on rbtree mutexes, instead of hand-rolling them
Alan T. DeKok [Thu, 18 Oct 2018 13:33:36 +0000 (09:33 -0400)]
more pthread fixes
Alan T. DeKok [Thu, 18 Oct 2018 13:31:41 +0000 (09:31 -0400)]
we always have pthread here, too
Alan T. DeKok [Thu, 18 Oct 2018 13:30:18 +0000 (09:30 -0400)]
we always have pthreads
Arran Cudbard-Bell [Wed, 17 Oct 2018 18:56:20 +0000 (14:56 -0400)]
Merge pull request #2332 from z-eos/master
Add files via upload
Alan T. DeKok [Wed, 17 Oct 2018 18:06:07 +0000 (14:06 -0400)]
relay on the top-level Makefile to build radsniff
we don't need to regenerate radsniff.mk
Alan T. DeKok [Wed, 17 Oct 2018 18:04:02 +0000 (14:04 -0400)]
Revert "this shouldn't be in git"
This reverts commit
759cab975259bfb3c0d407742d0414c0c762a50c .
Arran Cudbard-Bell [Wed, 17 Oct 2018 17:59:36 +0000 (13:59 -0400)]
Stupid OpenSSL compatibility fixes
Alan T. DeKok [Wed, 17 Oct 2018 17:53:04 +0000 (13:53 -0400)]
this shouldn't be in git
Alan T. DeKok [Wed, 17 Oct 2018 17:51:39 +0000 (13:51 -0400)]
OpenSSL can die in a fire.
They deprecated one API. And then added another one, which does
the same thing, but has a different name.
Alan T. DeKok [Wed, 17 Oct 2018 17:30:26 +0000 (13:30 -0400)]
Revert "add transitive linking for pre-requisite libraries"
This reverts commit
20d9dd6aae81569b07a83dc3b9ecd0e764a69773 .
Alan T. DeKok [Wed, 17 Oct 2018 17:30:25 +0000 (13:30 -0400)]
Revert "link only to the server library"
This reverts commit
5545374b06d2ef86271d882508abfa8e5c9133f1 .
Alan T. DeKok [Wed, 17 Oct 2018 17:30:22 +0000 (13:30 -0400)]
evert "simplify linking"
This reverts commit
f38c372bbe3302246ec94565f01c31c677f90c49 .
Alan T. DeKok [Wed, 17 Oct 2018 17:30:15 +0000 (13:30 -0400)]
Revert "simplify linking here, too"
This reverts commit
5abf2974fd04b62ee2dc94702f0dbab4481c07ae .
Alan T. DeKok [Wed, 17 Oct 2018 17:20:00 +0000 (13:20 -0400)]
simplify linking here, too
Alan T. DeKok [Wed, 17 Oct 2018 16:55:34 +0000 (12:55 -0400)]
simplify linking
Alan T. DeKok [Wed, 17 Oct 2018 16:52:39 +0000 (12:52 -0400)]
link only to the server library
which in turn links to libfreeradius-util
Alan T. DeKok [Wed, 17 Oct 2018 16:30:55 +0000 (12:30 -0400)]
add transitive linking for pre-requisite libraries
so that if we have:
radiusd -> server.a - > util.a
we don't need to link radiusd explicitly against util.a
For this step, we just add massive amounts of things to the
compiler flags. The next step is to start pruning the build
system of libraries
Arran Cudbard-Bell [Wed, 17 Oct 2018 16:50:55 +0000 (12:50 -0400)]
Initialise options
Arran Cudbard-Bell [Wed, 17 Oct 2018 14:07:33 +0000 (10:07 -0400)]
Add support for PCRE2_COPY_MATCHED_SUBJECT
As per ticket here: https://bugs.exim.org/show_bug.cgi?id=2323
Alan T. DeKok [Wed, 17 Oct 2018 14:06:43 +0000 (10:06 -0400)]
clang scan nonsense
Zeus Panchenko [Wed, 17 Oct 2018 10:58:23 +0000 (13:58 +0300)]
Add files via upload
attributes added, minor fix of objectClass
Zeus Panchenko [Wed, 17 Oct 2018 09:59:53 +0000 (12:59 +0300)]
Add files via upload
duplicates are removed
Alan T. DeKok [Tue, 16 Oct 2018 19:13:41 +0000 (15:13 -0400)]
disallow struct[n]
which means that we don't really care about the size of the
structs.
On encoding, if we run out of room, we just truncat the data
as with every other data type.
On decoding, if there's too little data, we just return what
data we can find. If there's too much data, we just ignore
the extra bits.
Alan T. DeKok [Tue, 16 Oct 2018 18:02:04 +0000 (14:02 -0400)]
try to quiet analyzer
Alan T. DeKok [Tue, 16 Oct 2018 17:32:33 +0000 (13:32 -0400)]
assertion for static analyzer
Alan T. DeKok [Tue, 16 Oct 2018 17:31:51 +0000 (13:31 -0400)]
fix static analysis warning
Alan T. DeKok [Tue, 16 Oct 2018 17:29:16 +0000 (13:29 -0400)]
clean up process_authenticate
in order to get rid of old attribute names
Alan T. DeKok [Tue, 16 Oct 2018 17:04:03 +0000 (13:04 -0400)]
remove more unused code
Alan T. DeKok [Tue, 16 Oct 2018 16:59:49 +0000 (12:59 -0400)]
we don't need Post-Auth-Type any more
the reply code has already been set by the time we're running
this module.