]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Alan T. DeKok [Tue, 12 Feb 2019 21:08:18 +0000 (16:08 -0500)]
note recent changes
Alan T. DeKok [Tue, 12 Feb 2019 21:07:17 +0000 (16:07 -0500)]
document new config
Alan DeKok [Tue, 12 Feb 2019 18:51:42 +0000 (13:51 -0500)]
Merge pull request #2481 from
2128506 /v3.0.x
V3.0.x
Arran Cudbard-Bell [Mon, 11 Feb 2019 04:26:52 +0000 (09:56 +0530)]
Expose more TLS options in MySQL driver
Availability of TLS and reconnect options in MySQL driver checked by using MySQL version instead of ifdefs
Alan T. DeKok [Thu, 7 Feb 2019 15:42:25 +0000 (10:42 -0500)]
check home server, not pool for virtual server.
We already picked a home server from the pool. And in some
cases, there's no pool, just a home server.
Alan T. DeKok [Tue, 5 Feb 2019 15:16:13 +0000 (10:16 -0500)]
add Module-Failure-Message here, too. Helps with #2464
Alan T. DeKok [Tue, 5 Feb 2019 13:35:13 +0000 (08:35 -0500)]
add Module-Failure-Message on failure. Addresses #2464
Alan T. DeKok [Fri, 1 Feb 2019 15:58:30 +0000 (10:58 -0500)]
bump the upper limit by a bit, to daily or so
Alan T. DeKok [Thu, 31 Jan 2019 19:38:18 +0000 (14:38 -0500)]
add 'M' as per v4 for packet usec
Alan T. DeKok [Thu, 31 Jan 2019 15:56:49 +0000 (10:56 -0500)]
only print open / close messages in debug mode
otherwise we spam the logs with things that aren't relevant.
The code already produces ERROR messages for error situations,
so that's fine.
Matthew Newton [Wed, 30 Jan 2019 21:56:14 +0000 (21:56 +0000)]
Merge pull request #2359 from mcnewton/v3.0.x
add crypt xlat
Matthew Newton [Thu, 22 Nov 2018 16:23:58 +0000 (16:23 +0000)]
add crypt xlat
Alan T. DeKok [Wed, 30 Jan 2019 20:51:45 +0000 (15:51 -0500)]
fixes for OpenSSL 1.1.1
Only send a single session ticket if the session cache is enabled,
and no session tickets if its not
Alan T. DeKok [Wed, 30 Jan 2019 14:36:51 +0000 (09:36 -0500)]
whitespace and minor fixes
It's important to keep consistent formatting, otherwise it
all goes to heck
Alan DeKok [Wed, 30 Jan 2019 14:31:37 +0000 (09:31 -0500)]
Merge pull request #2429 from maximumG/2395-sqlippool-allocate-clear_v3
issue 2395 sqlippool allocate clear v3
Alan T. DeKok [Tue, 29 Jan 2019 12:41:36 +0000 (07:41 -0500)]
more notes
Alan DeKok [Tue, 29 Jan 2019 12:34:27 +0000 (07:34 -0500)]
Merge pull request #2439 from restena-sw/patch-1
clarify which CA certs to include when using certificate_file without ca_file
Stefan Winter [Tue, 29 Jan 2019 12:26:33 +0000 (13:26 +0100)]
clarify which CA certs to include when using certificate_file without ca_file
Sending the root serves no useful purpose; the client has this root CA installed and verifies against that trust base, or it doesn't. No amount of sending it as part of the untrusted cert data will make the root CA any more or less trusted.
The previous wording was confusing for eduroam admins and consequently we see many who send a superfluous *root* certificate with their server cert. This typically costs another 1-2 roundtrips for no real purpose.
The new wording is more explicit that *intermediate* CAs are a good idea, but no root.
MaximumG [Mon, 28 Jan 2019 07:46:45 +0000 (08:46 +0100)]
Add support for timeout on sqlippool allocate_clear
Add allocate_clear_timeout support to sqlippol
Alan T. DeKok [Mon, 28 Jan 2019 13:33:39 +0000 (08:33 -0500)]
if not otherwise specified, set maximum TLS version to 1.2.
If TLS 1.4 is defined, we just use TLS 1.2. As we have no idea
what to do for TLS 1.4.
If TLS 1.3 is defined, we use TLS 1.2. Because the standards
for EAP-TLS are not finished yet. Much less those for TTLS and
PEAP.
Alan T. DeKok [Sun, 27 Jan 2019 16:25:57 +0000 (11:25 -0500)]
keep just the license text
Alan T. DeKok [Sun, 27 Jan 2019 14:47:57 +0000 (09:47 -0500)]
added dictionary for RFC 8045
Note that it has 3 attributes, all TLVs. Unlike other dictionaries,
these attributes share the same sub-TLV space. Since we can't
represent that in v3, we just gave them different names.
Alan T. DeKok [Sun, 27 Jan 2019 14:47:21 +0000 (09:47 -0500)]
update copyright to 2019
Alan T. DeKok [Sun, 27 Jan 2019 14:45:56 +0000 (09:45 -0500)]
add LICENSE to CC BY 4.0
and fix various formatting issues
Alan T. DeKok [Fri, 25 Jan 2019 12:43:02 +0000 (07:43 -0500)]
notes on cipher_server_preference being likely wrong
Alan T. DeKok [Thu, 24 Jan 2019 19:33:31 +0000 (14:33 -0500)]
config is 'const'
Alan DeKok [Sat, 26 Jan 2019 13:01:50 +0000 (08:01 -0500)]
Merge pull request #2419 from alejandro-perez/patch-1
Complete and improve realm module documentation
Alejandro Perez [Fri, 25 Jan 2019 16:24:31 +0000 (17:24 +0100)]
Complete and improve realm module documentation
Josef Vybíhal [Thu, 24 Jan 2019 11:30:10 +0000 (12:30 +0100)]
Add attribute number 14 (HPE-Port-MA-Port-Mode)
<Attribute profile="in out" type="Unsigned32" name="HPE-Port-MA-Port-Mode" id="14">
<ValidValues>
<ValidValue enumOrdinal="0" value="User-Based"/>
<ValidValue enumOrdinal="1" value="Port-Based"/>
</ValidValues>
</Attribute>
Alan T. DeKok [Tue, 22 Jan 2019 18:25:27 +0000 (13:25 -0500)]
note recent changes
Alan T. DeKok [Tue, 22 Jan 2019 18:23:40 +0000 (13:23 -0500)]
check for Vendor-Specific only if vendor==0
Alan T. DeKok [Tue, 22 Jan 2019 15:21:04 +0000 (10:21 -0500)]
remove sql_fields
Lots of code, and nothing ever used it.
Matthew Newton [Tue, 22 Jan 2019 14:02:13 +0000 (14:02 +0000)]
Add latest-alpine Dockerhub tag, too
Fixes #2306
Alan T. DeKok [Mon, 21 Jan 2019 20:00:27 +0000 (15:00 -0500)]
only do clear if it's configured
Alan DeKok [Mon, 21 Jan 2019 17:56:48 +0000 (12:56 -0500)]
Merge pull request #2394 from alejandro-perez/patch-3
Use tabs for indentation
Alejandro Perez [Mon, 21 Jan 2019 15:42:24 +0000 (16:42 +0100)]
Use tabs for indentation
Alan DeKok [Mon, 21 Jan 2019 12:37:26 +0000 (07:37 -0500)]
Merge pull request #2393 from alejandro-perez/patch-2
Be a bit more verbose in trustrouter.c
Alan DeKok [Mon, 21 Jan 2019 12:29:51 +0000 (07:29 -0500)]
Merge pull request #2392 from alejandro-perez/patch-1
Make realm examples be more generic
Alejandro Perez [Mon, 21 Jan 2019 11:53:28 +0000 (12:53 +0100)]
Be a bit more verbose in trustrouter.c
Add realm name associated to trust router queries in debug messages
Alejandro Perez [Mon, 21 Jan 2019 11:48:46 +0000 (12:48 +0100)]
Make realm examples be more generic
Current examples were very specific referring to Painless Security and janet.
Use more generic realms.
Alan T. DeKok [Sat, 19 Jan 2019 17:26:18 +0000 (12:26 -0500)]
note recent changes
Arran Cudbard-Bell [Sat, 19 Jan 2019 13:26:00 +0000 (20:26 +0700)]
Merge pull request #2391 from mcnewton/v3fixes
Tidying for next release
Matthew Newton [Sat, 19 Jan 2019 00:41:21 +0000 (00:41 +0000)]
make lintian happy on debian stretch
Matthew Newton [Sat, 19 Jan 2019 03:04:16 +0000 (03:04 +0000)]
enable more hardening options on debian
Matthew Newton [Fri, 18 Jan 2019 23:14:08 +0000 (23:14 +0000)]
consistent formatting
Matthew Newton [Fri, 18 Jan 2019 23:13:27 +0000 (23:13 +0000)]
typos
Matthew Newton [Fri, 18 Jan 2019 21:52:54 +0000 (21:52 +0000)]
update some copyrights
Alan T. DeKok [Fri, 18 Jan 2019 16:43:18 +0000 (11:43 -0500)]
notes on PSK, certs, and TLS 1.3
Alan T. DeKok [Fri, 18 Jan 2019 16:14:32 +0000 (11:14 -0500)]
note recent changes
Alan T. DeKok [Fri, 18 Jan 2019 16:08:03 +0000 (11:08 -0500)]
Fix code to note that PSK + certs is forbidden by the spec
Alan T. DeKok [Fri, 18 Jan 2019 15:38:05 +0000 (10:38 -0500)]
disable TLS 1.3 for EAP-FAST
Alan T. DeKok [Thu, 17 Jan 2019 21:48:50 +0000 (16:48 -0500)]
push free coa until after child has returned
Alan T. DeKok [Thu, 17 Jan 2019 15:22:25 +0000 (10:22 -0500)]
Note recent changes
Alan T. DeKok [Thu, 17 Jan 2019 15:20:23 +0000 (10:20 -0500)]
typo
Alan T. DeKok [Thu, 17 Jan 2019 13:25:27 +0000 (08:25 -0500)]
perhaps a better way of working around the issue.
Only warn about the OpenSSL bug if we're using TLS1.3.
Alan T. DeKok [Wed, 16 Jan 2019 17:05:28 +0000 (12:05 -0500)]
free request->coa on cancel, too
Alan T. DeKok [Wed, 16 Jan 2019 16:39:23 +0000 (11:39 -0500)]
disable TLS 1.3 for OpenSSL >1.1, AND PSK AND certs
because OpenSSL doesn't support it. Should help with #2390
Alan T. DeKok [Wed, 16 Jan 2019 16:22:13 +0000 (11:22 -0500)]
move "check for connection down" to coa_retransmit
as it is called from multiple places, it should check for
error conditions that apply to it.
Alan T. DeKok [Wed, 16 Jan 2019 16:20:26 +0000 (11:20 -0500)]
add CANCELLED state to request state machine.
This is administratively cancelled, to distinguish it from a
"finished normally" state. The main difference is that for
"finished normally", we send CoA packets. On cancellation,
we do not.
Alan T. DeKok [Wed, 16 Jan 2019 16:00:43 +0000 (11:00 -0500)]
don't free request->coa on a timer
the timer is hit multiple times, and not just at max_request_time
Alan T. DeKok [Wed, 16 Jan 2019 11:58:35 +0000 (06:58 -0500)]
this hasn't been in the server for a while
Alan T. DeKok [Tue, 15 Jan 2019 21:45:18 +0000 (16:45 -0500)]
note recent changes
Alan T. DeKok [Tue, 15 Jan 2019 19:29:26 +0000 (14:29 -0500)]
request->home_server may be NULL
Alan T. DeKok [Tue, 15 Jan 2019 19:21:04 +0000 (14:21 -0500)]
don't print two error messages
Alan T. DeKok [Tue, 15 Jan 2019 19:08:52 +0000 (14:08 -0500)]
Fix timers with CoA. May help with #2308 and #2328
The maximum time for CoA packets should be enforced from
when the CoA packet started, OR from `mrd`. Further, the
max time calculation shouldn't touch `request->delay`, because
that is used by `coa_retransmit()`
Alan T. DeKok [Tue, 15 Jan 2019 19:08:13 +0000 (14:08 -0500)]
fix problems with previous commit
Alan T. DeKok [Tue, 15 Jan 2019 14:18:27 +0000 (09:18 -0500)]
and more...
Alan T. DeKok [Tue, 15 Jan 2019 13:34:50 +0000 (08:34 -0500)]
remove more spurious commits
Alan T. DeKok [Tue, 15 Jan 2019 12:53:06 +0000 (07:53 -0500)]
remove merge text
Alan T. DeKok [Tue, 15 Jan 2019 12:37:07 +0000 (07:37 -0500)]
fix typos From #2296
Alan T. DeKok [Tue, 15 Jan 2019 12:27:24 +0000 (07:27 -0500)]
typo. Fixes #2357
Alan T. DeKok [Tue, 15 Jan 2019 12:12:49 +0000 (07:12 -0500)]
note recent changes
Alan DeKok [Tue, 15 Jan 2019 11:52:14 +0000 (06:52 -0500)]
Merge pull request #2387 from mcnewton/hintsfix
Fix crash processing hints when using Fall-Through
Matthew Newton [Tue, 15 Jan 2019 10:54:14 +0000 (10:54 +0000)]
Remove unnecessary pointer
radius_pairmove can change request->packet->vps in some situations
meaning request_pairs is left dangling at an old address.
This restores the v2 "Fall-Through = Yes" behaviour, too (i.e.
match on the current version of the request, not the original
incoming packet attributes).
Fixes #2380
Matthew Newton [Tue, 15 Jan 2019 10:11:27 +0000 (10:11 +0000)]
Clarify documentation on how hints works
Arran Cudbard-Bell [Tue, 15 Jan 2019 09:47:53 +0000 (16:47 +0700)]
Merge pull request #2386 from alejandro-perez/v3.0.x
Fix typo when verifying psk_identity and psk_password status
Alejandro Perez [Tue, 15 Jan 2019 09:46:46 +0000 (10:46 +0100)]
Fix typo when verifying psk_identity and psk_password status
Arran Cudbard-Bell [Tue, 15 Jan 2019 02:42:16 +0000 (09:42 +0700)]
Fix merge issue
Arran Cudbard-Bell [Tue, 15 Jan 2019 02:30:36 +0000 (09:30 +0700)]
Fix relatively severe memory leak in EAP-PWD
# Conflicts:
# src/modules/rlm_eap/types/rlm_eap_pwd/rlm_eap_pwd.c
Alan T. DeKok [Mon, 14 Jan 2019 21:19:44 +0000 (16:19 -0500)]
allow certs && PSK.
Also document this in the default config
Alan T. DeKok [Mon, 14 Jan 2019 16:39:01 +0000 (11:39 -0500)]
more descriptive error message
Alan T. DeKok [Mon, 14 Jan 2019 16:36:11 +0000 (11:36 -0500)]
rearrange checks and make the error messages more descriptive
Alan T. DeKok [Mon, 14 Jan 2019 13:43:56 +0000 (08:43 -0500)]
change test to soft fail if there's no session
in SSL init isn't finished, it's a soft fail if SSL_get_session()
doesn't return a session.
Alan T. DeKok [Sun, 13 Jan 2019 00:10:59 +0000 (19:10 -0500)]
bump copyright
Alan T. DeKok [Fri, 11 Jan 2019 12:22:29 +0000 (07:22 -0500)]
note recent changes
Alan DeKok [Fri, 11 Jan 2019 12:06:39 +0000 (07:06 -0500)]
Merge pull request #2384 from alejandro-perez/v3.0.x
Add libssl1.1 as patched for Heartbleed
Arran Cudbard-Bell [Fri, 11 Jan 2019 10:47:33 +0000 (05:47 -0500)]
Merge pull request #2383 from SearchLightNZ/sqlippool_skip_locked
Include SKIP LOCKED queries as options for sqlippool, some notes, and some tidy up.
Alejandro Perez [Thu, 10 Jan 2019 14:48:02 +0000 (15:48 +0100)]
Add libssl1.1 as patched for Heartbleed
Nathan Ward [Fri, 11 Jan 2019 03:23:44 +0000 (16:23 +1300)]
Add a note that ORDER BY RANDOM() is very slow.
This first gets all the free IPs, then randomises them, then selects the first one.
Nathan Ward [Fri, 11 Jan 2019 03:14:13 +0000 (16:14 +1300)]
Comment out sqlippool PostgreSQL allocate_find query, so only one allocate_find query is uncommented, mirroring MySQL and Oracle queries configs.
Nathan Ward [Fri, 11 Jan 2019 03:13:35 +0000 (16:13 +1300)]
Add SKIP LOCKED variations of sqlippool allocate_find queries as options.
Arran Cudbard-Bell [Tue, 8 Jan 2019 10:22:14 +0000 (18:22 +0800)]
Prevent crash in EAP-Fast if CHAP-Challenge is not the length we expect
Alan T. DeKok [Sun, 6 Jan 2019 14:55:03 +0000 (09:55 -0500)]
remove references to v3.1.x, and replace with master
Alan T. DeKok [Mon, 31 Dec 2018 15:57:06 +0000 (10:57 -0500)]
clarify message
Alan T. DeKok [Sun, 30 Dec 2018 13:40:10 +0000 (08:40 -0500)]
close when connect fails
Alan T. DeKok [Sun, 30 Dec 2018 13:20:52 +0000 (08:20 -0500)]
free empty nodes, too
Alan T. DeKok [Fri, 28 Dec 2018 14:43:45 +0000 (09:43 -0500)]
note recent changes
Alan T. DeKok [Fri, 28 Dec 2018 14:37:57 +0000 (09:37 -0500)]
don't artificially truncate debug output. Fixes #2381
The buffer size for VPs being printed should be as long as
the attribute name, plus spaces, plus operator, plus quotes,
plus encoded content. In the worst case, where a printable
string contains all binary data, it will be 4x the length of
the string. So we'll need ~64 + 6 + 4*247, or a bit more than
1K of room.
That's pathological, so we don't care so much about it.
Instead, increase the size of the buffer so that common cases
aren't truncated.
Note also that we print out the *full* string, even if the
encoded version is truncated to 257 octets. There's currently
no good way to expose that, unfortunately. We would need to
add a callback to the *encoding* routine which took a print
function, and could then pass the *encoded* length to the print
function.
Alan T. DeKok [Thu, 27 Dec 2018 21:42:22 +0000 (16:42 -0500)]
make it dependent on max_request_time
Alan T. DeKok [Thu, 27 Dec 2018 21:40:54 +0000 (16:40 -0500)]
set timer for free of dynamic clients
instead of magic in client_free() where it's not really needed,
put the timer into listen.c.