]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Thu, 16 Aug 2018 16:57:17 +0000 (12:57 -0400)]
add fd_set API
Alan T. DeKok [Thu, 16 Aug 2018 16:52:05 +0000 (12:52 -0400)]
remove debug message
Alan T. DeKok [Thu, 16 Aug 2018 16:45:52 +0000 (12:45 -0400)]
do connect() here, instead of the UDP module
Alan T. DeKok [Thu, 16 Aug 2018 16:45:44 +0000 (12:45 -0400)]
move set name out of instantiate for RADIUS
and stop doing connect() here
Alan T. DeKok [Thu, 16 Aug 2018 14:07:42 +0000 (10:07 -0400)]
move set name out of instantiate for RADIUS
Alan T. DeKok [Thu, 16 Aug 2018 13:58:51 +0000 (09:58 -0400)]
move set name out of instantiate
Alan T. DeKok [Thu, 16 Aug 2018 13:53:55 +0000 (09:53 -0400)]
move set name out of instantiate
Alan T. DeKok [Thu, 16 Aug 2018 13:52:46 +0000 (09:52 -0400)]
use get_name if available for connection name
Alan T. DeKok [Thu, 16 Aug 2018 13:52:33 +0000 (09:52 -0400)]
rearrange documentation
Alan T. DeKok [Thu, 16 Aug 2018 13:34:04 +0000 (09:34 -0400)]
verify connect_set is available
and only call instantiate if it exists
Alan T. DeKok [Thu, 16 Aug 2018 13:28:38 +0000 (09:28 -0400)]
this needs to be socket_instance for now
Alan T. DeKok [Thu, 16 Aug 2018 12:49:55 +0000 (08:49 -0400)]
move set name to open function
Alan T. DeKok [Wed, 15 Aug 2018 18:39:40 +0000 (14:39 -0400)]
start of separation of config && thread data for network IO
we now have "app_io_instance" which is the parsed configuration
data.
And we also have "socket_instance", which is the data that changes
at run-time. e.g. socket FD, list of dynamic clients, statistics,
etc.
This commit just adds the API capabilities. The next few commits
should start separating the responsibilities of the various pieces
of information.
Alan T. DeKok [Mon, 13 Aug 2018 17:54:09 +0000 (13:54 -0400)]
rename socket_add to listen_add
Alan T. DeKok [Mon, 13 Aug 2018 14:34:17 +0000 (10:34 -0400)]
one ifdef is needed
Alan T. DeKok [Mon, 13 Aug 2018 12:55:40 +0000 (08:55 -0400)]
move tls_free() to header definition
so that there are fewer ifdef's in the code
Alan T. DeKok [Sat, 11 Aug 2018 01:27:35 +0000 (21:27 -0400)]
remove tls_free if not available. Fixes #2272
Alan T. DeKok [Fri, 10 Aug 2018 15:32:34 +0000 (11:32 -0400)]
we need to track our own parent for get_inst
Alan T. DeKok [Fri, 10 Aug 2018 15:15:56 +0000 (11:15 -0400)]
remove calls to talloc_parent()
as it's fugly
Alan T. DeKok [Fri, 10 Aug 2018 15:13:57 +0000 (11:13 -0400)]
move each client to it's own talloc context
because each client can have many thousands of packets associated
with it, and we want to avoid O(N) issues in talloc
Alan T. DeKok [Fri, 10 Aug 2018 11:14:25 +0000 (07:14 -0400)]
Revert "show full log"
This reverts commit
d8514645242f2bf96cb0c2032e6fdbb0604943ba .
Alan T. DeKok [Fri, 10 Aug 2018 01:33:04 +0000 (21:33 -0400)]
parsed is not referenced
referencing an item shouldn't prevent it from being parsed
Alan T. DeKok [Thu, 9 Aug 2018 19:14:19 +0000 (15:14 -0400)]
show full log
Alan T. DeKok [Thu, 9 Aug 2018 18:50:35 +0000 (14:50 -0400)]
typo
Arran Cudbard-Bell [Thu, 9 Aug 2018 18:48:16 +0000 (14:48 -0400)]
Don't leave hanging message in non-debug builds
Alan T. DeKok [Thu, 9 Aug 2018 18:35:01 +0000 (14:35 -0400)]
print out number of network / workers on error
Alan T. DeKok [Thu, 9 Aug 2018 18:34:47 +0000 (14:34 -0400)]
print out error message on failure
Alan T. DeKok [Thu, 9 Aug 2018 18:02:21 +0000 (14:02 -0400)]
don't call talloc_parent()
instead, cache ef->el for FDs
Alan T. DeKok [Thu, 9 Aug 2018 18:01:51 +0000 (14:01 -0400)]
use 'nr' instead of 'rc'
and pass 'nr' when adding network commands
Alan T. DeKok [Thu, 9 Aug 2018 17:22:52 +0000 (13:22 -0400)]
cache event list in event_timer_t
because talloc_parent() is O(N) in the number of objects
in the talloc context. Which is slow as heck, especially
if you're looping over events and deleting them one by one.
At that point, it becomes O(N^2), and all kinds of bad things
start happening.
Alan T. DeKok [Thu, 9 Aug 2018 17:20:48 +0000 (13:20 -0400)]
revert debug thread code
it's not the problem
Alan T. DeKok [Wed, 8 Aug 2018 19:24:12 +0000 (15:24 -0400)]
go to error instead of exit
Alan T. DeKok [Wed, 8 Aug 2018 18:34:31 +0000 (14:34 -0400)]
be more stringent on allow varargs
Alan T. DeKok [Wed, 8 Aug 2018 18:33:44 +0000 (14:33 -0400)]
typo
Alan T. DeKok [Wed, 8 Aug 2018 18:16:20 +0000 (14:16 -0400)]
might as well check all API calls, too
Alan T. DeKok [Wed, 8 Aug 2018 18:11:58 +0000 (14:11 -0400)]
try to catch possible threading issues?
Alan T. DeKok [Wed, 8 Aug 2018 15:25:33 +0000 (11:25 -0400)]
bump up max_workers
and ensure that the limit in the main config is the same
as in the network code
Alan T. DeKok [Wed, 8 Aug 2018 15:15:20 +0000 (11:15 -0400)]
allow any name2 for thread section
Alan T. DeKok [Wed, 8 Aug 2018 14:56:20 +0000 (10:56 -0400)]
use num_workers here, too
Alan T. DeKok [Wed, 8 Aug 2018 14:26:25 +0000 (10:26 -0400)]
set 1 spare
Alan T. DeKok [Wed, 8 Aug 2018 13:57:33 +0000 (09:57 -0400)]
use num_workers in pool instead of min/max/spare servers
We've got to clean up the connection pool code, too...
Alan T. DeKok [Wed, 8 Aug 2018 13:53:20 +0000 (09:53 -0400)]
Revert "this might work around the assertion"
This reverts commit
78387637e068d1a8c34dce24c0147851a6da8ae5 .
nope.
Alan T. DeKok [Wed, 8 Aug 2018 13:26:49 +0000 (09:26 -0400)]
delete old documentation and write new docs
for network / worker threads
Alan T. DeKok [Tue, 7 Aug 2018 20:00:00 +0000 (16:00 -0400)]
qualify ms == 1ms
Alan T. DeKok [Tue, 7 Aug 2018 19:35:53 +0000 (15:35 -0400)]
note that some things aren't thread-safe in the future
Alan T. DeKok [Tue, 7 Aug 2018 19:35:08 +0000 (15:35 -0400)]
allow for finer-grained selection of worker stats
Alan T. DeKok [Tue, 7 Aug 2018 19:34:20 +0000 (15:34 -0400)]
be more careful about valid UTF-8 names
and validate the name in the table
Alan T. DeKok [Tue, 7 Aug 2018 19:11:04 +0000 (15:11 -0400)]
remove "expiry_time IS NULL" from update query. Fixes #2262
Oracle and PostgreSQL already had that change made.
Alan T. DeKok [Tue, 7 Aug 2018 18:23:14 +0000 (14:23 -0400)]
check for NULL variable
Alan T. DeKok [Tue, 7 Aug 2018 18:21:44 +0000 (14:21 -0400)]
this might work around the assertion
Alan T. DeKok [Tue, 7 Aug 2018 18:19:12 +0000 (14:19 -0400)]
add assert for pre-existing ev
Alan T. DeKok [Tue, 7 Aug 2018 15:25:54 +0000 (11:25 -0400)]
rename again
Alan T. DeKok [Tue, 7 Aug 2018 15:20:07 +0000 (11:20 -0400)]
don't write zero bytes, write a space instead
Alan T. DeKok [Tue, 7 Aug 2018 15:09:33 +0000 (11:09 -0400)]
be a little more specific about the output
Alan T. DeKok [Tue, 7 Aug 2018 15:00:29 +0000 (11:00 -0400)]
add more worker stats
Alan T. DeKok [Tue, 7 Aug 2018 14:50:06 +0000 (10:50 -0400)]
track histogram data
Alan T. DeKok [Tue, 7 Aug 2018 14:33:57 +0000 (10:33 -0400)]
run different subcommands, not the same one...
Alan T. DeKok [Tue, 7 Aug 2018 14:33:32 +0000 (10:33 -0400)]
format output a little more nicely
Arran Cudbard-Bell [Tue, 7 Aug 2018 14:40:55 +0000 (10:40 -0400)]
Merge pull request #2271 from nchaigne/4.0.x-201808-dhcpclient-no-xlat
dhcpclient - xlat is not supported
nchaigne [Tue, 7 Aug 2018 14:35:25 +0000 (16:35 +0200)]
dhcpclient - xlat is not supported
Apply the same processing as performed in radclient to input value pairs
Alan T. DeKok [Tue, 7 Aug 2018 14:09:30 +0000 (10:09 -0400)]
remame for clarity
Alan T. DeKok [Tue, 7 Aug 2018 14:00:30 +0000 (10:00 -0400)]
fixed stats, and added "serviced"
so that we know when the worker last serviced the event loop
Alan T. DeKok [Tue, 7 Aug 2018 14:00:08 +0000 (10:00 -0400)]
added '*' in certain places as a primitive "for" loop
Alan T. DeKok [Tue, 7 Aug 2018 13:18:09 +0000 (09:18 -0400)]
remove struct signing
Alan T. DeKok [Mon, 6 Aug 2018 19:52:55 +0000 (15:52 -0400)]
mark unused variable
Alan T. DeKok [Mon, 6 Aug 2018 12:58:15 +0000 (08:58 -0400)]
this needs to be integer64
even tho it's a bit field
Alan T. DeKok [Mon, 6 Aug 2018 15:29:58 +0000 (11:29 -0400)]
remove testing code
Arran Cudbard-Bell [Mon, 6 Aug 2018 18:22:33 +0000 (14:22 -0400)]
Don't parent the signals event list from the global context
It's a singleton structure so easy to manage, and the memory needs to be written to at runtime so it doesn't work with mprotect()
Alan T. DeKok [Mon, 6 Aug 2018 15:03:55 +0000 (11:03 -0400)]
typo
Alan T. DeKok [Mon, 6 Aug 2018 14:27:19 +0000 (10:27 -0400)]
more debugging
Alan T. DeKok [Mon, 6 Aug 2018 14:01:53 +0000 (10:01 -0400)]
code to verify before sending
Arran Cudbard-Bell [Sun, 5 Aug 2018 23:05:33 +0000 (19:05 -0400)]
Print when we're protecting/unprotecting global memory
Arran Cudbard-Bell [Sun, 5 Aug 2018 23:05:18 +0000 (19:05 -0400)]
Bad arithmetic in talloc_page_aligned_pool
Alan T. DeKok [Sun, 5 Aug 2018 15:14:46 +0000 (11:14 -0400)]
re-sign after changing member values
Alan T. DeKok [Sun, 5 Aug 2018 15:02:06 +0000 (11:02 -0400)]
fix API to use talloc ctx
Alan T. DeKok [Sun, 5 Aug 2018 14:29:33 +0000 (10:29 -0400)]
use local radmin_ctx && free it on stop
Arran Cudbard-Bell [Sun, 5 Aug 2018 13:26:15 +0000 (09:26 -0400)]
Typo
Arran Cudbard-Bell [Sun, 5 Aug 2018 12:57:53 +0000 (08:57 -0400)]
Ignore leaks in libkqueue, there a fix in the library
Alan T. DeKok [Sun, 5 Aug 2018 12:51:54 +0000 (08:51 -0400)]
sign / verify clients that we've copied
Arran Cudbard-Bell [Sun, 5 Aug 2018 05:01:01 +0000 (01:01 -0400)]
Turn on leak detection
Arran Cudbard-Bell [Sun, 5 Aug 2018 04:59:16 +0000 (00:59 -0400)]
Add option to pre-allocate and protect global memory
Set FR_GLOBAL_POOL to use this feature
Arran Cudbard-Bell [Sun, 5 Aug 2018 04:58:30 +0000 (00:58 -0400)]
Various cleanups
Arran Cudbard-Bell [Sat, 4 Aug 2018 02:11:42 +0000 (22:11 -0400)]
We have a function for that...
Arran Cudbard-Bell [Fri, 3 Aug 2018 23:27:07 +0000 (19:27 -0400)]
It's not a uint8_t
Alan T. DeKok [Fri, 3 Aug 2018 20:32:47 +0000 (16:32 -0400)]
add place-holder for commands
so that if we add 100 commands of "... module foo", the help
text will only show help for "... module STRING ..."
which is much more polite than showing hundreds of possible
commands
Alan T. DeKok [Fri, 3 Aug 2018 20:23:49 +0000 (16:23 -0400)]
Don't allow radmin when checking the config.
Alan T. DeKok [Fri, 3 Aug 2018 19:08:51 +0000 (15:08 -0400)]
cast to pointer before assigning it
Alan T. DeKok [Fri, 3 Aug 2018 18:16:32 +0000 (14:16 -0400)]
don't write 0 bytes to SSL. Fixes #2270
Alan T. DeKok [Fri, 3 Aug 2018 16:40:52 +0000 (12:40 -0400)]
no longer have this
Arran Cudbard-Bell [Fri, 3 Aug 2018 16:43:28 +0000 (12:43 -0400)]
Add function for allocating talloc pools on page boundaries
Alan T. DeKok [Fri, 3 Aug 2018 14:57:32 +0000 (10:57 -0400)]
add "show client config" command
Alan T. DeKok [Fri, 3 Aug 2018 14:54:53 +0000 (10:54 -0400)]
update "word" when matching a data type
Alan T. DeKok [Thu, 2 Aug 2018 21:03:52 +0000 (17:03 -0400)]
comment out tab expand
until such time as we fix it for recent API changes
given that we're not using it, and instead are using
command_complete... it would be better to just delete this code
Alan T. DeKok [Thu, 2 Aug 2018 21:03:43 +0000 (17:03 -0400)]
fix for new API
Alan T. DeKok [Thu, 2 Aug 2018 20:20:15 +0000 (16:20 -0400)]
fix up so that adding intermediate names work
all fr_cmd_table_t now have to have `.name`
where we want to add a qualifier, e.g. `network X`, the table
has to have `.add_name = true,` added to it.
and fix `,dup` in network.c && worker.c
Alan T. DeKok [Thu, 2 Aug 2018 20:19:37 +0000 (16:19 -0400)]
track sockets by number, too
Alan T. DeKok [Thu, 2 Aug 2018 20:18:29 +0000 (16:18 -0400)]
skip the argc that we matched.
Corresponding to commit
cdd3d1260
Alan T. DeKok [Thu, 2 Aug 2018 17:28:30 +0000 (13:28 -0400)]
add "name" fr_cmd_table_t
so that we can specify dynamic names, e.g. "network 0 ..."
Alan T. DeKok [Thu, 2 Aug 2018 17:27:47 +0000 (13:27 -0400)]
remember that we skipped a word
Alan T. DeKok [Thu, 2 Aug 2018 16:51:50 +0000 (12:51 -0400)]
skip the word we matched