need help with wifi

[Samuel Ortiz]

Hi Sean,

On Mon, Apr 21, 2008 at 11:13:43AM -0700, Sean McNeil wrote:
> Hi Samuel,
> 
> Samuel Ortiz wrote:
> >On Sat, Apr 19, 2008 at 01:31:09PM -0700, Sean McNeil wrote:
> >  
> >>I have a wpa_supplicant built with wext. I invoke it as
> >>
> >>wpa_supplicant -Dwext -ieth0 -c/etc/wpa_supplicant.conf
> >>
> >>my wpa_supplicant.conf is
> >>
> >>----------------------------------------------------------------
> >>ctrl_interface=/tmp/wpa_supplicant
> >>
> >>network={
> >>   ssid="Guest"
> >>   key_mgmt=NONE
> >>}
> >>-----------------------------------------------------------------
> >>    
> >So, your AP is fully open, right ?
> >  
> No, this is just a dummy entry. I don't have a network of this type. 
> Mine are WPA and WEP protected.
> 
> >>with the latest andy git, I get the following error when it continuously 
> >>scans:
> >>
> >>AR6000 scan complete: 
> >>0                                                        
> >>WMI event ID : 0x1004, len = 301 too big for IWEVCUSTOM (max=256)
> >>AR6000 scan complete: 
> >>0                                                        
> >>    
> >Ok, I would need 2 console outputs from you:
> >1) The output of "iwlist wlan0 scan", before running wpa_supplicant.
> >2) The output of "wpa_supplicant -dd -Dwext -ieth0 
> >-c/etc/wpa_supplicant.conf"
> >Note the additional -dd compared to your command line. This will give us
> >plenty of debug from wpa_supplicant.
> >  
> Thanks, but I don't have iwlist available. I am using software that is 
> interfacing directly to the device via. IOCTL (I believe. Haven't gotten 
> into the guts of that part of the software yet).
> 
> >>Can someone help point me to who is at fault here? Is it the client 
> >>talking to wpa_supplicant, wpa_supplicant, or the kernel who is setting 
> >>up a 301 byte buffer? The message is printed out by the kernel, but I 
> >>don't know where the buffer originates.
> >>    
> >It's a buffer we received from the firmware, containing BSSI data. For some
> >reason (most likely a huge number of IEs), it's too big to be sent to
> >userspace.
> >
> >One more question: when you run wpa_supplicant, you keep on getting those
> >scan completew events, but it never actually associates with your AP ? Did
> >I understand you right ?
> >  
> Yes, I'm not even trying to associate with an AP. I'm just trying to get 
> the list of available APs to display. 
There's an existing library for that, it's called libiw, see:
http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/Tools.html

This library implements the userspace side of the linux wireless
extensions, and the wireless tools (iwconfig, iwlist, etc...) are
built on top of it. Since the gta02 wifi driver heavily relies on
those extensions for its kernel<->userspace communication, I think
implementing yet another ioctl library for fetching scanning
results somehow sounds like re-inventing the wheel to me.

> So it would seem things are broken 
> at the kernel level? I should think the buffer would be split up and 
> sent piecemeal. I did notice once that it returned an SSID to me in a 
> MAC address form (hex bytes separated by :). Is that correct? Seems like 
> the SW stack is expecting an integer.
Things are not broken at the kernel level. The error message you're seeing
is a failure to send a custom wireless event. It surely happens when
you're starting a scan, but it shouldn't prevent you from getting
your scan results.
Could you please describe which ioctls you're calling for starting
a scan and then fetching the san results ?

Cheers,
Samuel.