FW: Strange Enterprise bug

Jukka Rissanen jukka.rissanen at linux.intel.com
Wed Sep 18 23:58:53 PDT 2013


Hi Thomas,

On 18.09.2013 18:49, Thomas Green wrote:
> I've had my qa guy come up with a strange bug when connecting to an enterprise network.  Perhaps it would be best to just forward his mail on how to reproduce the issue.  I hope this is enough information.  If you need to contact me privately for more information my address is Tgreen2 at Sorenson dot com.

Did you had this 1ef0b07ebd81f2f7057f60c46b16a45e52396572 commit in your 
ConnMan version? It should have fixed the WPA-EAP issue you reported 
earlier. ConnMan 1.18 has that fix.


>
> From: Isaac Goodrich
> Sent: Thursday, September 12, 2013 2:20 PM
> To: Thomas Green
> Subject: Strange Enterprise bug
>
>
> When re-connecting to an Enterprise network after previously being connected to it, there is an occasional failure to re-connect. If you try to re-connect again, there is a prompt for credentials. If bad credentials are given, it appears as though connection was successful and the status shows online, but network connectivity is not present.
>
>
> 1.       Successfully connect to Enterprise network
>
> 2.       Disconnect
>
> 3.       Re-connect
>
> 4.       Repeat steps 2-3 until "Operation aborted" error occurs. Seems to occur when the re-connect is done too quickly after disconnecting.
>
> 5.       Try to re-connect (this time, you are prompted for username and password)
>
> 6.       Enter invalid credentials
>
> When this happens, there is no failure given and it appears as though it was successful. The status shows online. The problem is that there is not a connection. There is no IP address and DNS requests don't come back. At this point, if I disconnect from the network and re-connect, things are successful again and there is a connection. This means that the original good credentials, which were cached, are still being used on subsequent connection attempts during the session.
>
> I can't think of what would cause the problem on the connection attempt when entering bad credentials after an "operation aborted" error. The AP does not return a failure, so that seems to indicate that the cached good credentials are being sent, which also explains Connman not returning an error. Something causes Internet not to work in this scenario. My opinion is that to fix the issue, the cached credentials should be deleted after a failure to connect, so that the new credentials that are entered are the ones that get sent.
>
> Isaac Goodrich
> Software Engineer in Test - nVP
> Sorenson Communications
>

Cheers,
Jukka




More information about the connman mailing list