reassociation problem

[Gingernut]

This problem also exists in the Tomato firmware but only in the ND versions, ND = New Broadcom Driver.

I did some testing with Tomato and found more or less the same problems. No security or WEP and reassosiation is more or less fine but with WPA/WPA2 and the problems start.

I was told that it is a Broadcom driver problem at binary level.

I'am not sure what driver version Gargoyle/OpenWRT are using ?

[Eric]

That's... interesting. This problem is only being observed on devices with atheros drivers, which are completely different. The brcm driver, though, is the old, closed 2.4 driver, though.

I didn't realize that Tomato had a version now that used the newer driver, though I do know that it has been stated multiple times that the open-source driver is not ready for prime-time.

[crisman]

So, this could indicate that problem is in hostapd because the old broadcom driver use nas, but new one use hostapd, same as atheros. But I don't know why dd-wrt firmware is not affected by this problem; maybe they found a workaround? I read the dd-wrt trac searching hostapd as keyword and I found something interesting:

on changeset 10646 they pass from hostapd 0.6.5 to 0.6.2 ( but message reports "update supplicant and hostapd"; this is very strange );

on changeset 11213 they switch to 0.6.5 ( but message report "use older version"; very very strange);

on changeset 11214 they go to 0.6.2 ( but message report "get back to newer version since we fixed now the driver (assoc wpa problem)") ahah we found something interesting: "assoc wpa problem".

on changeset 11215 they switch back to 0.6.5


I think the messages are "traslated" by 1 changeset for such an error. But the important thing is they told about a driver fix for the assoc wpa problem.
Now we must find this driver fix, or trying to use hostapd 0.6.2 and verify if problem is also present.

Can you help me?

Best regards

[Gingernut]

I didn't realize that the problem was only Atheros based or at least only reported by Atheros users.
I haven't even got round to trying the Gargoyle firmware yet, but do hope to in the near future, so I don't know if the problem exists on the Broadcom side but I suppose not.

The Tomato creater decided to support some newer Broadcom based routers and these have updated cpu, wireless chip and switch that have to use these newer drivers and cannot use the older version because of lack of driver support.

Edit: DDWRT I think uses highly modified drivers as it lets you use wpa2 encription in bridge, WDS and client bridge modes.

[crisman]

Edit: DDWRT I think uses highly modified drivers as it lets you use wpa2 encription in bridge, WDS and client bridge modes.

client bridge is present also in Gargoyle and you can use what encryption you want ( I think ).
I think the hostapd package is the same as openwrt.

[crisman]

I compiled an openwrt revision with hostapd 0.6.2, luci, wpa_supplicant and pppoe-relay.

Can anyone try this and verify if bug is also present?

If you are interested tell me and I'll upload it.


Best regards

[Eric]

If you send me a link I can download it and test this tonight.

[crisman]

This is the link:

Code: Select all

http://www.easy-share.com/1906571671/openwrt_hostapd0.6.2.tar.gz

In the tar.gz you will find vmlinux.lmza, root.squashfs, root.jffs2-64k e ubntns2. So files are for fonera or nanostation2.

Let me know test results.

Thanks

[Eric]

It doesn't seem to work. I see the same symptoms as before.

Actually, what I'm seeing differs a little from what you report: I see it working for a few minutes before the connection dies and then I need to explicitly tell the computer to reassociate with the AP. I see exactly the same with the latest builds of Gargoyle.

[crisman]

this is not good

However, can you post the log to see if messages are the same?


Thanks