NEW Package: freeRADIUS 2.x

Nachtfalke

Long startup time of freeradius2 with pfsense 2.0.1

I have a WLAN system based on pfSense running in a small hotel, where user access is handled by the captive portal of pfSense using freeradius2 for authentication. I am getting some complaints about not being able to get connected to the internet. The captive portal is enabled all the time.

I can to reproduce this "login" problem in the following situations:

(1) After booting, when all services are already running, pfSense does not redirect to the login page of the captive portal and the HTML request just times out.
(2) A few times, after waiting for more than 2 minutes, all of a sudden the login page is shown and subsequently everything works as expected.
(3) When changing parameters of the freeradius2 configuration (probably freeradius2 is restarting in the background), the login screen doesn't show up again.
(4) If the captive portal is enabled without using any authentication, the login screen always appears (that's why I think it could be related to freeradius2 and not the captive portal itself).

Is this a known behavior of freeradius2? In case it just takes a couple of minutes (but why does it take so long?) until the logins work, I can live with that, as regular operation doesn't require frequent restarts or rebooting. Otherwise there might be a problem, at least for my system.

Hi,

don't know why CP isn't showing up with some delay. I don't think it has something to do with freeradius2 because if you enter a wrong RADIUS IP in CP then it is the same as freeradius not responding. But I am not sure at this.

There is a longer (repeating) startup time after reinstallation of freeradius2 package because of of the recreation of the config files. There is no way around.
But you are right - the startup of freeradius2 isn't really fast but I don't know how to speed this up.

Restart of freeradius is very fast if you:
(1) add/delete a user
(2) add/delete a client
In syslog you can see the output

Hmmm…if you find some more information when or why this is happening, please let me know. I will think about that "problem".

Fesoj

Nachtfalke,

I'll keep an eye on the issue. Let me first collect more information to find out under what precise circumstances the login problem occurs. I am pretty sure that it is not the server itself as "radtest <user><user pw="">…" works immediately with a very short response time after the server is marked as running inside the gui. At the same time the client login fails (by not showing the cp login page).

There could be a delay, but it could also be a random error in the way different components interact. I need to work out a suitable test matrix to nail down the problem (booting, restarting services, login immediately or after some time period). Maybe this will give a clearer picture...</user></user>

Fesoj

Long startup time of freeradius2 with pfsense 2.0.1

Meanwhile I rebooted the pfSense2 box in question serveral times && checked the internet access after waiting at least 5 minutes. I was able to login all the time. Therefore I think that the inability to login is just a startup timing problem.

I haven't checked the exact details, but if Status->Services reports a running service, it probably just means that a pid file exists, but that doesn't mean that the service is fully available to the outside world. My box uses an Atom 510 CPU and PowerD is enabled. Besides the CP the services Squid, SquidGuard (with a rather large blacklist) and Snort (a few rules) are enabled and especially Snort uses initially a lot of CPU time after the service has started up. The combination of resource hogs and a relatively weak CPU seem to be the cause of the reported problem.

Nachtfalke

@Fesoj:

Long startup time of freeradius2 with pfsense 2.0.1

Meanwhile I rebooted the pfSense2 box in question serveral times && checked the internet access after waiting at least 5 minutes. I was able to login all the time. Therefore I think that the inability to login is just a startup timing problem.

I haven't checked the exact details, but if Status->Services reports a running service, it probably just means that a pid file exists, but that doesn't mean that the service is fully available to the outside world. My box uses an Atom 510 CPU and PowerD is enabled. Besides the CP the services Squid, SquidGard (with a rather large blacklist) and Snort (a few rules) are enabled and especially Snort uses initially a lot of CPU time after the service has started up. The combination of resource hogs and a relatively weak CPU seem to be the cause the reported problem.

The startup of freeradius2 is probably not the best way. The problem is, that some files needs to be created/changed and after this freeradius needs a restart. I can not replace the restart of freeradius after some functions because I need the restart when changing something from GUI. I am running freeradius2 package on a xeon processor and that's probably why I do not have problems with weak CPU power ;o)

PS: Sometimes the green service arrow isn't representing the actual state. If you have logging enabled you should see how often freeradius is restarting in the syslog.

Fesoj

Nachtfalke,

when you say "…and after this freeradius needs a restart" does that mean the freeradius process gets killed and restarted quite frequently, especially after new users have been added?

Nachtfalke

@Fesoj:

Nachtfalke,

when you say "…and after this freeradius needs a restart" does that mean the freeradius process gets killed and restarted quite frequently, especially after new users have been added?

When adding/deleting a users it is absolutly essential to restart the freeradius service. But it is not killed (killall -9) it gets restarted (/usr/local/etc/rc.d/radiusd onerestart)
It is further essential if you add a client, and interface, change eap settings, change SQL or LDAP settings. It is simple essential after all changes you are doing. There ist no way around - this is the way freeradius2 works :)

Fesoj

I just looked at radiusd.c and before doing anything else gets done the configuration files are read and this is the only place where read_mainconfig() gets called (and not at later times again triggered by a signal as is done in dnsmasq). The executable is small and there are only a few small local files to read, so restarting shouldn't take a lot of time. Also I do not see any problems when new users are added, only after starting the box (I still need to verify under which conditions restarting of freeradius and/or other services cause the login problems).

Tomorrow I'll restart the pfSense box again a few times and try to access the internet within the first 30 seconds after the box has signalled that it is ready. Meanwhile I also have a second box prepared with an older single core AMD CPU, so I can study the behavior of freeradius2 without having the Snort and the Squid stuff installed (just in case these packages slow down the entire system) and then see what happens if I add these packages again.

Currently I tend to think that an Atom based system needs probably about 2 minutes until it is usable if many resource hungry packages are installed.

Fesoj

Maybe the problem of not being able to log in is not at all related to freeRADIUS2. The PortalAuth log has the following entries whenever I restart the Squid/SquidGuard daemons:

Feb 29 20:03:45 squid[15693]: Squid Parent: child process 16200 exited with status 0
Feb 29 20:03:48 squid[30849]: Squid Parent: child process 31376 started
Feb 29 20:04:03 squid[31376]: The url_rewriter helpers are crashing too rapidly, need help!
Feb 29 20:04:03 squid[30849]: Squid Parent: child process 31376 exited due to signal 6
Feb 29 20:04:06 squid[30849]: Squid Parent: child process 53720 started

I searched for the url_rewriter problem, but I did not find a real solution. This message seems to show up quite frequently, so it could just be a bogus message.

Fesoj

Somewhere else somebody suggested to check the db files of SquidGuard when SquidGuard is having problems. I did it and got:

[2.0.1-RELEASE][admin@pfsense2.bridge]/var/db/squidGuard/blk_BL_aggressive(40): /usr/local/bin/db42/db_verify domains.db
db_verify: Page 1: out-of-order key at entry 2
db_verify: Page 1: out-of-order key at entry 4
db_verify: Page 1: out-of-order key at entry 6
db_verify: Page 1: out-of-order key at entry 10
…
db_verify: DB->verify: domains.db: DB_VERIFY_BAD: Database verification failed

The files urls.db are ok, but all the domains.db files failed. Downloading the black list again and letting the system rebuild the database does not help.

Does somebody know whether one needs to worry about this?

Nachtfalke

@Fesoj:

Somewhere else somebody suggested to check the db files of SquidGuard when SquidGuard is having problems. I did it and got:

[2.0.1-RELEASE][admin@pfsense2.bridge]/var/db/squidGuard/blk_BL_aggressive(40): /usr/local/bin/db42/db_verify domains.db
db_verify: Page 1: out-of-order key at entry 2
db_verify: Page 1: out-of-order key at entry 4
db_verify: Page 1: out-of-order key at entry 6
db_verify: Page 1: out-of-order key at entry 10
…
db_verify: DB->verify: domains.db: DB_VERIFY_BAD: Database verification failed

The files urls.db are ok, but all the domains.db files failed. Downloading the black list again and letting the system rebuild the database does not help.

Does somebody know whether one needs to worry about this?

I got the same error when checking another domains.db on my machine. But I don't think this is really related to freeradius2 and this thread ;)

Fesoj

I got the same error when checking another domains.db on my machine. But I don't think this is really related to freeradius2 and this thread

In my case I called db_verify on an system with SquidGuard running. This might not be the best idea. Before I'll continue I need to do some mandatory reading on Berkeley DBs. So far I know that db_verify doesn't acquire any locks, but SquidGuard might. In that case the error messages could be bogus.

… but now it's time to go home.

saye

Hi
I am running 2.0.1 64 bit and this issue of not starting freeradius2 is still exist.

I really would like to know is there any fix of this bug in close future?

I know that the great pfsense team are working on 2.1 release but some features of pfsense may excel brilliant over other UTMs and will broaden its popularity.
Thus as a matter of fact I should say that this feature of pfSense( authenticating via a radius server for large number of users ) and also synchronizing pfsense and authenticating pfsense users via windows radius and then adding some parameters( such as time quota) that windows radius will miss is a great idea that I swear will be one of best features of pfsense.

Cheers to one of the best open source projects of all time.

tanx

Nachtfalke

@saye:

Hi
I am running 2.0.1 64 bit and this issue of not starting freeradius2 is still exist.

I really would like to know is there any fix of this bug in close future?

(…)

Don't know about any bug that freeradius2 is not starting !?!

You should explain that more in detail:

1.) When it is not starting - after pfsense reboot ?
2.) What does syslog say - did you enable logging ?

freeradius2 package is configured by default that you ONLY need to add an interface and the the service is running. Try and test if this is working!
Don't use the old freeradius package and the freeradius2 package together!

Fesoj

Tomorrow I'll restart the pfSense box again a few times and try to access the internet within the first 30 seconds

I restarted the box a couple of times (having verified that the configuration of the system worked) and tried to login via CP/freeRADIUS2 after about 30s. There were no login problems at all.

Now it looks as if some kind of previous misconfiguration was responsible for the login problems.

The only freeRADIUS related messages were

Mar 1 14:17:48 radiusd[4350]: rlm_radutmp: Login entry for NAS pfsense2 port 2 wrong order
Mar 1 14:17:48 radiusd[4350]: rlm_radutmp: Login entry for NAS pfsense2 port 2 wrong order

and

Mar 1 14:18:08 radiusd[4350]: rlm_radutmp: Logout for NAS pfsense2 port 2, but no Login record
Mar 1 14:18:08 radiusd[4350]: rlm_radutmp: Logout for NAS pfsense2 port 2, but no Login record

but I guess these are without further consequences.

Nachtfalke

The rlm_radutmp problems are related to a known bug/mishandling of CaptivePortal.
But it does not make any problems if you do not use Simultaneous-Use = 1. Higher values are working.

http://doc.pfsense.org/index.php/FreeRADIUS_2.x_package#KNOWN_BUGS

qbik

Hi I am using this package and created a group with 2 attributes: Max-All-Session := 600 (5 min) Simultaneous-Use := 1. I've enabled the noresetcounter on radiusd.conf, users on that group are able to authenticate ok, however they do not get disconnected after 600 sec have past. CP is enabled, interim updates enabled, authenticate users every minute is enabled. I have the same setup on another pf sense server with freeradius installed but not with this package and all is working ok on that one. I am using MySQL on both.

Thanks

saye

I 've tried it on 2.0 release and it's working
but on 2.0.1 I 've tested it and it appears that 'modules' directory is missing

I tried: radiusd -X
and it said:
: Error reading directory /usr/local/etc/raddb/modules/: No such file or directory

I's a strange problem.
Thank you in advance

marcelloc

@saye:

I 've tried it on 2.0 release and it's working
but on 2.0.1 I 've tested it and it appears that 'modules' directory is missing

I tried: radiusd -X
and it said:
: Error reading directory /usr/local/etc/raddb/modules/: No such file or directory

Both installs are i386 or x64?
Can you try a package reinstall? Both 2.0 and 2.0.1 use the same package from freebsd.

Nachtfalke

@qbik:

Hi I am using this package and created a group with 2 attributes: Max-All-Session := 600 (5 min) Simultaneous-Use := 1. I've enabled the noresetcounter on radiusd.conf, users on that group are able to authenticate ok, however they do not get disconnected after 600 sec have past. CP is enabled, interim updates enabled, authenticate users every minute is enabled. I have the same setup on another pf sense server with freeradius installed but not with this package and all is working ok on that one. I am using MySQL on both.

Thanks

Hi,
do you use freeradius2 package on both pfsense or not ? That's not clear for me.

the attribute: Max-All-Session is not a predefined attribute in GUI. So you probably created this attribute in your mysql database ? And if you use mysql as database for accounting then you need to use the sqlcounter.conf file. This file is NOT supported from GUI. You need to check and perhaps edit this file from shell.

I don't how it works with mxsql but if I use the ../raddb/modules/counter module (without external database) then this modul is ONLY running on accounting-stop packets, interim-updates do not work. So make sure that it does with mysql.

re-authenticate every minute should be enabled on CP so that the user gets disconnected. that's right what you did. You should enabled "Disable acct_unique" in FreeRADIUS -> Settings to make accounting work properly.

@saye:

I 've tried it on 2.0 release and it's working
but on 2.0.1 I 've tested it and it appears that 'modules' directory is missing

I tried: radiusd -X
and it said:
: Error reading directory /usr/local/etc/raddb/modules/: No such file or directory

I's a strange problem.
Thank you in advance

Modules directory will be created when installing freeradius2 package. It is part of the default freeradius installation. Nothing changed there. Probably some problem while reinstallation.

Check if this folder exists - if not - reinstall package.

I tested with 2.0.1 and it is working.

Nachtfalke

Updates pkg v1.6.6:

Added: Possibility to sync to a remote host which has a different GUI protocol/port as the sync master
Bumped package to RC1 - there seems to be no reportet bug related to this package and at the moment there are no feature improvements planned.

Known bugs:

When using "stop/start accounting on CP then "Amount of Time/Amount of Traffic" isn't working correctly.
http://redmine.pfsense.org/issues/2164
When using CP + RADIUS + Vouchers and "reauthenticate every minute" is enabled then CP sends the voucher as username to RADIUS. This causes RADIUS to disconnect the "user/voucher" because of an unknown/wrong "username".
http://redmine.pfsense.org/issues/2155
When stop/start accounting on CP is enabled than the syslog shows many "wrong order" or "Login found bot no logout detected". This seems to not affect the usage of RADIUS but it is not 100% correct.
http://redmine.pfsense.org/issues/2143