Captive portal ignoring MACs in latest version and allowing all machines access
-
I understand - I'm just repeating myself a lot (close to rambling perhaps).
You and I use the same code. I really want see this bug. Seeing is is resolving it.If I install the latest pfSense on a separate machine and import this configuration, the problem exists also on the new machine. Perhaps I could sanitize the configuration file and post it here. Perhaps you can see something I can't. But I know it can be reproduced. It's not a complicated router-- one WAN, one LAN. No hot backup.
-
@h2professor said in Captive portal ignoring MACs in latest version and allowing all machines access:
Perhaps I could sanitize the configuration file and post it here
Start with the entire <captiveportal> </captiveportal> xml dump.
Don't post it here : drop it on a pastebin.org and send me the link.What packages are installed on this setup ?
-
@Gertjan said in Captive portal ignoring MACs in latest version and allowing all machines access:
@h2professor said in Captive portal ignoring MACs in latest version and allowing all machines access:
Perhaps I could sanitize the configuration file and post it here
Start with the entire <captiveportal> </captiveportal> xml dump.
I'll do this a bit later this morning. I think I already sent you everything except for the MAC entries.
Can you send me your email address? I can send you these files directly. Send me an email h2professor@gmail.comWhat packages are installed on this setup ?
arping*
bandwidthd
cron*
darkstat*
FTP_Client_Proxy
notes*- can be uninstalled because we never use it... installed by the client's request years ago.
-
Notes and cron cron are totally inoffensive,
Same thing for arpping, I just tested that one.
FTP_Client_Proxy : I thought that FTP was abandoned .... but I'll install it anyway.
bandwidthd and darkstat : never used them, but I'm pretty sure they do "mess" with the IP stack ....
-
FTP_Client_Proxy : I thought that FTP was abandoned .... but I'll install it anyway.
We have no choice on this one. There are legacy devices in the organization that have to communicate with legacy FTP servers externally. I'm sure there's a workaround, but I can't believe this would suddenly be a problem after using it for years.
bandwidthd and darkstat : never used them, but I'm pretty sure they do "mess" with the IP stack ....
These could be uninstalled temporarily (darkstat permanently) but again, been using them for years with no problem.
edit: I just removed Darkstat and had no effect on the problem. -
@h2professor said in Captive portal ignoring MACs in latest version and allowing all machines access:
been using them for years with no problem.
A recent update ?
Packages are not static.
For example, it happened that FreeRadius was upgraded and breaking my portal access because "there was a small glitch" in the newer version.I don't use darkstat or bandwidthd so I don't know if they were upgraded, which could be a potential reason.
I'll install them anyway.Btw : if this setup isn't patched or changed manually, you could take a backup (config file) of the system, and remove these packages for testing purposes. re-installing them afterwards, importing the backup file and a reboot would take care of things.
I'll report back Monday, have to leave now. Weekend ;)
-
Did you resolve this? We are running into a problem that seems similar:
- We have incrementally upgraded from pfsense 1.x to 2.4.4.p3
- We have MAC addresses (but also allowed IP addresses)
- Setting bandwidth limits on the MAC addresses works, but the captive portal does not block blocked MAC addresses
- All users are able to use the network without the captive portal
- DNS works and is being done by pfSense
- The firewall allows http/https access
However, one thing is different: removing all MAC address listings does NOT fix the problem. The captive portal continues not to work.
I can generate the results of the ipfw commands above but do not what I am looking for.
-
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
Did you resolve this? We are running into a problem that seems similar
Yes, that does seem similar. We did not get this solved with the latest version. We ended up rolling back to a mirror-drive backup dated October 7, 2017 and it's working. (The configuration is 100% identical other than the software version.) We're not going to apply any updates until we figure out how to get it working on the latest version. Nobody in this forum believes there's a real problem, so it's somewhat comforting that someone else is working the same problem, although I do apologize that it had to be you.
Coincidentally my client is shutting down for a company-wide getaway this coming week so I'll be able to focus my attention on getting the most recent version working, if it's possible. I might experiment with 2.5.x experimental releases if all else fails.
I will post my findings in this forum.
Thank you
-
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
removing all MAC address listings does NOT fix the problem
Question: are you using a custom Captive Portal login page? Or the stock firewall login page?
Thank you -
@h2professor It is a custom login page.
Unlike your situation we do ask for a username and password, using the local database.
-
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
It is a custom login page
If it's no trouble, may I suggest that you delete your custom page and see if the problem resolves without it?
In a non-production setup last July (with the same configuration) we ended up in a situation where everything was working until we added the custom page. If this happens with yours as well, then we might narrow down the problem a bit.
Thanks -
No, when I delete the custom login page and use the default I get the same behavior.
-
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
I get the same behavior
okay thank you for trying. We'll proceed with our testing here and will share anything that might be useful.
Cheers -
@h2professor said in Captive portal ignoring MACs in latest version and allowing all machines access:
Nobody in this forum believes there's a real problem,
You had side-long talks in this topic about your problem and say nobody believes you? I find that hard to believe.
As the last of this was talked about in June/July, did you @h2professor ever try to re-create the steps manually to "create" the problem you have?
After reading through, I was always stuck with "and then I re-imported my whole/part of my config and the problem was back". Is it possible to re-trace the steps for the problem to appear? As @Gertjan said: seeing is (no, not believing) half way to fix a potential bug. But as this always re-appeared after updating/re-importing the config I'm wondering if it's a problem with the configuration as it is.
Also did anyone try and open a bug report at the pfsense redmine and link this thread? Might be helpful for hunting down a potential bug!
https://redmine.pfsense.org/projects/pfsense/issues -
@JeGr said in Captive portal ignoring MACs in latest version and allowing all machines access:
I find that hard to believe
How ironic.
We tried everything that was suggested, but the client ran out of patience for the problem, so we were forced to use the last working cold mirror to get it back online.
Here's a summary:
It worked prior to 2.4.4-p3. It has worked for years.
We upgraded to p3: it no longer works.
We did a clean installation of p3: confirmed captive portal works as advertised in a basic installation on the hardware.
We imported the original configuration: no longer works.
We did a clean installation of the older 2.4.4-p1 and imported configuration: worked just fine.
We tried dozens of suggestions, including removing packages, reinstalling, cutting sections out of the configuration, etc.We had to conclude that the problem is in version 2.4.4-p3 despite people insisting that "no changes were made to the captive portal section" bla bla bla. Hence, it's not a believable problem.
I have a week with the client out of their building so I have time to figure this out, if it is possible.
-
Well, I solved the problem in my installation, but I expect it won't help @h2professor . We have some allowed IPs in "Allowed IP addresses." One of our local IPs was listed there. However, that dialog allows you to set a subnet mask, and the IP in that subnet had been set to /24, not /32. So it was allowing the entire subnet through!
In the process of finding this, I rebuilt my captive portal:
- I created a dummy interface and assigned the old captive portal there.
- I made the configurations the same for the new portal.
- I added all of the IP addresses and hostnames (but idiot that I was, I inadvertently used a /32 for the offending IP in my new build, and not a /24 like in the old one)
- I added the allowed MAC addresses one by one
- I re-added my custom login page
and sure enough the new portal worked, but the old one didn't. (I tested the new captive portal after adding each component, hoping to find the one that broke the build.)
I also exported the captive portal config for these two captive portals and compared them, but I was not able to see any significant differences.
The only thing that occurred to me was whether there might be a duplicate MAC address in the list? Due to a copy and paste error when transcribing entries I thought I had one. The current pfSense interface will not let me add a duplicate MAC address, but maybe some old one did?
-
It is not quite true to say that the XML for the two captive portal backups were identical. There were some extra lines in the old one that were not present in the new one:
<radius_protocol></radius_protocol> <redirurl></redirurl> <radiusip></radiusip> <radiusip2></radiusip2> <radiusip3></radiusip3> <radiusip4></radiusip4> <radiusport></radiusport> <radiusport2></radiusport2> <radiusport3></radiusport3> <radiusport4></radiusport4> <radiusacctport></radiusacctport> <radiuskey></radiuskey> <radiuskey2></radiuskey2> <radiuskey3></radiuskey3> <radiuskey4></radiuskey4> <radiusvendor>default</radiusvendor> <radiussrcip_attribute>wan</radiussrcip_attribute>(
redirurlis actually in both).I notice that
radiussrcip_attributeis set towanin my listing and set tolanin yours. I tried changing mine toopt1(my captive portal interface) in my backup file and restoring, but it made no difference. -
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
I solved the problem in my installation
That's excellent work, thank you very much. We're going to try that. It's something we haven't considered. I'll report back in a couple of days.
Cheers -
@theworkingcentre said in Captive portal ignoring MACs in latest version and allowing all machines access:
I notice that radiussrcip_attribute is set to wan in my listing and set to lan in yours. I tried changing mine to opt1 (my captive portal interface) in my backup file and restoring, but it made no difference.
When the config file version moved from 18.6 to 18.7 (august 2018), that's several updates in the past, we're 19.1 now - "radiussrcip_attribute" was created and set to 'wan' as de default value.
It's a radius related setting, and if radius isn't used it isn't important.
<captiveportal> <cpzone1> <zone>cpzone1</zone> <descr><![CDATA[cpzone1]]></descr> <zoneid>2</zoneid> <interface>opt1</interface> ...... <radiussrcip_attribute>opt1</radiussrcip_attribute> ...... </cpzone1> </captiveportal>I'm using FreeRadius - my portal is running on OPT1.
-
@h2professor said in Captive portal ignoring MACs in latest version and allowing all machines access:
We had to conclude that the problem is in version 2.4.4-p3 despite people insisting that "no changes were made to the captive portal section" bla bla bla. Hence, it's not a believable problem.
People don't "insist" anything. My post also wasn't meant as an insult or attack, I was simply wondering why you wrote that. The devs simply said in another thread, that they did check the portal code and there was no commit or anything else belonging to the portal code in between -p1 and -3. So it either has to be a dependency like e.g. radius or some other thing that links into the portal like IP lists etc. what @theworkingcentre wrote.
I also am not "ironic" but I find it quite a bit "unfair" to write sth like "nobody believes me" after page-long tries to help you debug the problem Also your list in "things we did do" lists "importing the configuration" EVERY time. So like I wrote above, how about simply debugging the thing with a fresh installation and configuring step by step the way it is now in your clients environment? Why do you insist of always importing the config.xml when it could very well be some thing that doesn't get imported the right way or perhaps was a problem with importing and updating this installation every time? I had small problems like that a few times. Rarely but they do exist. Changes in config format and old configs that have some duplicate lines or sth along those lines.
So why not trying one of this things instead of being sarcastic yourself and accuse us for doing "nothing"? At the end of the day we are simply other users and/or private persons here. This is not the official Netgate pfSense bug tracking support system.
And as much as I would like to see you/us succeed in tackling that problem: If your client is that sensitive and you have that narrow a margin in errors, wouldn't it be a better job to buy support for that device and ask the devs to have a look / fix it in your client's setup instead of getting angry with us for trying to help you? Or at least before going the support route submit that case with all details as a an official bug in the bugtracker? Sorry, I'm just trying to help here. And I'm not related or employed by Netgate after all. :)
Best wishes!
