]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Fri, 20 Dec 2019 13:26:51 +0000 (08:26 -0500)]
add "show profile", and don't allow "profile on" multiple times
Alan T. DeKok [Thu, 19 Dec 2019 21:36:20 +0000 (16:36 -0500)]
avoid cursors where they're not necessary
Arran Cudbard-Bell [Fri, 20 Dec 2019 02:08:31 +0000 (18:08 -0800)]
Don't hardcode the src ip address
The DHCPv4 code has multiple ways of discovering the correct src IP. Hardcoding the src IP to 127.0.0.1 means sending responses will fail will obscure errors, which is frustrating.
The server works out-of-the-box biding to '*'.
Arran Cudbard-Bell [Fri, 20 Dec 2019 00:45:14 +0000 (16:45 -0800)]
Quiet compiler
Arran Cudbard-Bell [Thu, 19 Dec 2019 23:24:34 +0000 (15:24 -0800)]
This is only an indication *IF* we're waiting
Arran Cudbard-Bell [Thu, 19 Dec 2019 21:57:45 +0000 (13:57 -0800)]
Always initialise priority
Arran Cudbard-Bell [Thu, 19 Dec 2019 21:53:26 +0000 (13:53 -0800)]
Simplify logic slightly in fr_event_corral
Arran Cudbard-Bell [Thu, 19 Dec 2019 21:52:34 +0000 (13:52 -0800)]
Add notes on why we call el->time() again
Arran Cudbard-Bell [Thu, 19 Dec 2019 21:17:08 +0000 (13:17 -0800)]
Need to pass GPERFTOOLS_LDFLAGS too
Alan T. DeKok [Thu, 19 Dec 2019 20:40:12 +0000 (15:40 -0500)]
reset time after doing significant work, and before running timers
Alan T. DeKok [Thu, 19 Dec 2019 20:36:00 +0000 (15:36 -0500)]
try to account for time sleeping
Alan T. DeKok [Thu, 19 Dec 2019 19:37:17 +0000 (14:37 -0500)]
return correct value from all situations
Alan T. DeKok [Thu, 19 Dec 2019 18:49:00 +0000 (13:49 -0500)]
if we get a timeout in kevent(), there MUST be timer events
Alan T. DeKok [Thu, 19 Dec 2019 18:45:55 +0000 (13:45 -0500)]
remove unused pre-event
it's only used if the worker is suppressing signals to the
network side. Since we always signal, this code isn't necessary
Alan T. DeKok [Thu, 19 Dec 2019 18:43:07 +0000 (13:43 -0500)]
only run pre_callbacks if the caller tells us to wait
only the network & worker threads use the pre callbacks. And,
the only thing they do is lower the wait time. So if we're told
to not wait, then don't bother asking caller if the timers
should be lowered.
Alan T. DeKok [Thu, 19 Dec 2019 18:42:52 +0000 (13:42 -0500)]
no need to check for a bool > 0. It is, if it's true
Alan T. DeKok [Thu, 19 Dec 2019 18:38:08 +0000 (13:38 -0500)]
try to simplify wait / timer logic
Alan T. DeKok [Thu, 19 Dec 2019 16:27:47 +0000 (11:27 -0500)]
notes on profile builds
Alan T. DeKok [Thu, 19 Dec 2019 16:24:12 +0000 (11:24 -0500)]
don't need a special profile build
Terry Burton [Thu, 19 Dec 2019 18:35:22 +0000 (18:35 +0000)]
Link radiusd with -lprofiler when gperftools is in use
Jorge Pereira [Fri, 13 Dec 2019 21:52:23 +0000 (18:52 -0300)]
Fix memory leak in rlm_python (Python 3.x)
We should release 'inst->module' in Python 3 too.
Alan T. DeKok [Thu, 19 Dec 2019 15:05:02 +0000 (10:05 -0500)]
handle is always set. Found by PVS-Studio
Alan T. DeKok [Thu, 19 Dec 2019 15:04:00 +0000 (10:04 -0500)]
move more assignments
Alan T. DeKok [Thu, 19 Dec 2019 14:48:17 +0000 (09:48 -0500)]
check for Access-Reject, not Access-Accept. Found by PVS-Studio
Alan T. DeKok [Thu, 19 Dec 2019 14:44:16 +0000 (09:44 -0500)]
check before dereferencing. Found by PVS-Studio
Alan T. DeKok [Thu, 19 Dec 2019 14:23:01 +0000 (09:23 -0500)]
check variable before dereferencing it. Found by PVS-Studio
Alan T. DeKok [Thu, 19 Dec 2019 14:18:49 +0000 (09:18 -0500)]
fix check for UTF-8 characters. Found by PVS-Studio
Alan T. DeKok [Thu, 19 Dec 2019 14:03:20 +0000 (09:03 -0500)]
add "set profile status on/off [filename]" command
Jorge Pereira [Wed, 18 Dec 2019 20:26:05 +0000 (17:26 -0300)]
The talloc_* options should be FR_TYPE_HIDDEN (#3217)
* Add new 'FR_TYPE_HIDDEN' type
It's helpful to set the developer options.
* The talloc_* options should be FR_TYPE_HIDDEN
* Skip the 'FR_TYPE_HIDDEN' options
Alan T. DeKok [Wed, 18 Dec 2019 19:31:41 +0000 (14:31 -0500)]
auth logs are INFO, not (by default) EMERGENCY
Alan T. DeKok [Wed, 18 Dec 2019 15:49:42 +0000 (10:49 -0500)]
add "to do" list for timers, too
New timers were added while running timers. Instead of
being added to the main heap, they were added to the
"in progress" list. We can now add them to the main
heap.
Doing it this way prevents the server from running
into an infinite loop when it takes a long time to
service FD events. The timer callback MAY add a new
timer which is in the past. So the above loop could
run forever.
Alan T. DeKok [Wed, 18 Dec 2019 14:57:12 +0000 (09:57 -0500)]
is broken on travis. Disabling until we track it down
Alan T. DeKok [Wed, 18 Dec 2019 14:50:15 +0000 (09:50 -0500)]
set min / max size for network ring buffers
Jorge Pereira [Wed, 18 Dec 2019 00:31:54 +0000 (21:31 -0300)]
Updated by 'make dictionary.format'
Alan T. DeKok [Wed, 18 Dec 2019 14:36:49 +0000 (09:36 -0500)]
use L_AUTH for authentication logs
Alan T. DeKok [Wed, 18 Dec 2019 14:25:26 +0000 (09:25 -0500)]
re-add L_AUTH
Arran Cudbard-Bell [Wed, 18 Dec 2019 05:48:11 +0000 (12:48 +0700)]
Need to use the 64bit format specifier
Arran Cudbard-Bell [Wed, 18 Dec 2019 05:46:02 +0000 (12:46 +0700)]
Redis doesn't like it when you pass the time in nanoseconds
Arran Cudbard-Bell [Wed, 18 Dec 2019 04:20:29 +0000 (11:20 +0700)]
vb_date was converted to microsecond reslution but rlm_cache wasn't updated to handle that
Arran Cudbard-Bell [Wed, 18 Dec 2019 03:46:24 +0000 (10:46 +0700)]
vb_date is microseconds since the epoch, i.e. an external time, and so SHOULD NOT HAVE OUR INTERNAL TIME SUBTRACTED FROM IT
Arran Cudbard-Bell [Wed, 18 Dec 2019 03:33:25 +0000 (10:33 +0700)]
...and here
Arran Cudbard-Bell [Wed, 18 Dec 2019 03:30:43 +0000 (10:30 +0700)]
Use the correct type
Alan T. DeKok [Wed, 18 Dec 2019 01:08:53 +0000 (20:08 -0500)]
Add missing field. Fixes #3215
Jorge Pereira [Tue, 17 Dec 2019 17:59:19 +0000 (14:59 -0300)]
Updated by 'make doc.check'
Jorge Pereira [Tue, 17 Dec 2019 19:26:54 +0000 (16:26 -0300)]
Update scripts/build/missing-raddb-mod-conf.txt db
Jorge Pereira [Tue, 17 Dec 2019 19:26:50 +0000 (16:26 -0300)]
Add capability to verify undocumented 'FR_CONF.*'
Jorge Pereira [Tue, 17 Dec 2019 19:30:46 +0000 (16:30 -0300)]
Verify the 'output' parameter
Jorge Pereira [Tue, 17 Dec 2019 19:22:09 +0000 (16:22 -0300)]
Update scripts/build/missing-xlat-doc.txt db
Alan T. DeKok [Tue, 17 Dec 2019 16:14:34 +0000 (11:14 -0500)]
don't debug print pairs twice
Alan T. DeKok [Tue, 17 Dec 2019 16:01:33 +0000 (11:01 -0500)]
move parsing of workers / threads into scheduler
Alan T. DeKok [Tue, 17 Dec 2019 15:47:16 +0000 (10:47 -0500)]
typo
Alan T. DeKok [Tue, 17 Dec 2019 15:01:57 +0000 (10:01 -0500)]
parse information for debugging
Alan T. DeKok [Tue, 17 Dec 2019 14:49:10 +0000 (09:49 -0500)]
pass "thread" CONF_SECTION to scheduler
and on to other things
Alan T. DeKok [Tue, 17 Dec 2019 14:24:41 +0000 (09:24 -0500)]
provisions for print out statistics
Alan T. DeKok [Tue, 17 Dec 2019 14:17:41 +0000 (09:17 -0500)]
API to dump stats to a log
Alan T. DeKok [Tue, 17 Dec 2019 14:17:25 +0000 (09:17 -0500)]
API to dump stats to a log
Alan T. DeKok [Tue, 17 Dec 2019 13:45:51 +0000 (08:45 -0500)]
move stats to stats structure
Arran Cudbard-Bell [Tue, 17 Dec 2019 12:38:41 +0000 (19:38 +0700)]
Not necessarily an error
Arran Cudbard-Bell [Tue, 17 Dec 2019 10:47:11 +0000 (17:47 +0700)]
Print out the actual errors instead of just ignoring them
Arran Cudbard-Bell [Tue, 17 Dec 2019 09:37:18 +0000 (16:37 +0700)]
Add sanity checks to make sure the channel we're signalling on is open
Arran Cudbard-Bell [Tue, 17 Dec 2019 09:36:52 +0000 (16:36 +0700)]
Fix worker name
Arran Cudbard-Bell [Tue, 17 Dec 2019 08:32:39 +0000 (15:32 +0700)]
Print login success/failure messages at the correct point
Arran Cudbard-Bell [Tue, 17 Dec 2019 08:32:13 +0000 (15:32 +0700)]
More information would be helpful
Arran Cudbard-Bell [Tue, 17 Dec 2019 08:04:06 +0000 (15:04 +0700)]
Missed fd delete
Arran Cudbard-Bell [Tue, 17 Dec 2019 04:33:50 +0000 (11:33 +0700)]
Remove dependencie on wbclient
Arran Cudbard-Bell [Tue, 17 Dec 2019 04:15:43 +0000 (11:15 +0700)]
Make wbclient optional
Arran Cudbard-Bell [Tue, 17 Dec 2019 04:04:27 +0000 (11:04 +0700)]
Move ready to process requests
Arran Cudbard-Bell [Tue, 17 Dec 2019 04:04:11 +0000 (11:04 +0700)]
Unused
Arran Cudbard-Bell [Tue, 17 Dec 2019 00:43:38 +0000 (07:43 +0700)]
Revert "Switch to atomics for shared fields"
This reverts commit
ee803e75e3c430dc2539ed38e9728f5425f7eeb2 .
# Conflicts:
# src/lib/io/channel.c
Arran Cudbard-Bell [Mon, 16 Dec 2019 20:35:50 +0000 (03:35 +0700)]
Don't populate TLS until we need it
Arran Cudbard-Bell [Mon, 16 Dec 2019 20:31:30 +0000 (03:31 +0700)]
Address possible weird interation with threads/pthread API
Alan T. DeKok [Mon, 16 Dec 2019 20:23:12 +0000 (15:23 -0500)]
make "enable skips" code to compile
still doesn't help tho. likely because other code also needs
to be updated to skip signalling
Alan T. DeKok [Mon, 16 Dec 2019 20:03:36 +0000 (15:03 -0500)]
remove get type abort.
the instruction could be any kind of unlang_*_t
Alan T. DeKok [Mon, 16 Dec 2019 20:01:37 +0000 (15:01 -0500)]
set track_duplicates flag in the listener
because the master_io handler can track duplicates, but perhaps
the underlying app_io doesn't need that.
Alan T. DeKok [Mon, 16 Dec 2019 19:42:43 +0000 (14:42 -0500)]
use correct variable
Alan T. DeKok [Mon, 16 Dec 2019 19:26:20 +0000 (14:26 -0500)]
add more decoding tests for options with 255 octets of data
Alan T. DeKok [Mon, 16 Dec 2019 19:05:15 +0000 (14:05 -0500)]
use atomic loads for reading atomic variables
Alan T. DeKok [Mon, 16 Dec 2019 19:04:56 +0000 (14:04 -0500)]
check both ends of the channel, not the same end twice
Alan T. DeKok [Mon, 16 Dec 2019 18:56:38 +0000 (13:56 -0500)]
cache compiled unlang
which saves almost 10% on the performance tests.
Alan T. DeKok [Mon, 16 Dec 2019 18:56:04 +0000 (13:56 -0500)]
allow caching of compiled instruction
Alan T. DeKok [Mon, 16 Dec 2019 18:54:27 +0000 (13:54 -0500)]
remove unused assignment
Alan T. DeKok [Mon, 16 Dec 2019 18:33:33 +0000 (13:33 -0500)]
return compiled instruction if the caller asks for it
Alan T. DeKok [Mon, 16 Dec 2019 18:28:50 +0000 (13:28 -0500)]
add API to push bare instructions into a request
which is faster than doing lookups in CONF_SECTIONS
Alan T. DeKok [Mon, 16 Dec 2019 18:21:54 +0000 (13:21 -0500)]
update reference for newer specification
Alan T. DeKok [Mon, 16 Dec 2019 18:19:16 +0000 (13:19 -0500)]
dhcpv4 options can be 257 octets long, not 255.
and move the "check for limit" code to decode option, not to
the callers.
Alan T. DeKok [Mon, 16 Dec 2019 03:14:16 +0000 (22:14 -0500)]
bump up the default sizes for ring buffers.
to 128K, which should be lots for most purposes
Alan T. DeKok [Mon, 16 Dec 2019 03:10:15 +0000 (22:10 -0500)]
error messages are good for debugging, but not for the hot loop
It turns out that hitting the limits of ring buffers happens
a lot in the load tests.
Alan T. DeKok [Mon, 16 Dec 2019 02:06:30 +0000 (21:06 -0500)]
set CLOEXEC on pipes, too
Alan T. DeKok [Mon, 16 Dec 2019 01:54:47 +0000 (20:54 -0500)]
allow bypassing / closing of all pipes and kqueues
if the two ends of the channel are in the same thread, just
run the callback immediately.
Arran Cudbard-Bell [Mon, 16 Dec 2019 17:52:08 +0000 (00:52 +0700)]
Don't set defaults which'll cause the server to fail to start
Arran Cudbard-Bell [Mon, 16 Dec 2019 17:34:08 +0000 (00:34 +0700)]
Not American
Arran Cudbard-Bell [Mon, 16 Dec 2019 15:12:43 +0000 (22:12 +0700)]
Use common log prefix for worker messages
Arran Cudbard-Bell [Mon, 16 Dec 2019 14:29:20 +0000 (21:29 +0700)]
Fix fr_value_box_cast_to_octets when dealing with non-numeric and non-ip types.
Matthew Newton [Mon, 16 Dec 2019 13:35:40 +0000 (13:35 +0000)]
don't try and build rlm_python if we can't configure it
Arran Cudbard-Bell [Mon, 16 Dec 2019 13:14:05 +0000 (20:14 +0700)]
Fix error code return from unit_test_module
Arran Cudbard-Bell [Mon, 16 Dec 2019 13:00:52 +0000 (20:00 +0700)]
Don't need to print this repeatedly
Arran Cudbard-Bell [Mon, 16 Dec 2019 13:00:30 +0000 (20:00 +0700)]
Doesn't need to be atomic
Arran Cudbard-Bell [Mon, 16 Dec 2019 11:37:11 +0000 (18:37 +0700)]
Use more specific error messages
Arran Cudbard-Bell [Mon, 16 Dec 2019 11:12:43 +0000 (18:12 +0700)]
Typo
Arran Cudbard-Bell [Mon, 16 Dec 2019 10:58:01 +0000 (17:58 +0700)]
Remove fr_worker_exit and strip fr_ prefix off local functions