Roaming to same BSSID as currently associated behaviour.
Ben Greear
greearb
Mon Nov 18 17:23:04 PST 2013
On 11/18/2013 02:54 PM, Jouni Malinen wrote:
> On Mon, Nov 18, 2013 at 02:32:27PM -0800, Ben Greear wrote:
>> On 11/18/2013 02:25 PM, Dan Williams wrote:
>>> On Sun, 2013-11-17 at 19:11 -0800, Ben Greear wrote:
>>>> But, what is the preferred method of roaming if the roam command is
>>>> just for testing purposes?
>>>
>>> Use one of the bgscan plugins and let the supplicant handle roaming
>>> itself based on signal strength. Or if it's a fullmac driver, the
>>> driver will handle roaming.
>>
>> For my purposes I want more control over it, so I guess the 'roam' cli
>> command will be best.
>
> If I've understood your use case correctly, it would have been very much
> in my "testing purposes" category.. ;-) It is fine to use the roam
> command, but you'll need to be more aware of the driver details (and
> especially, cfg80211 behavior in this case) when using it. Please take a
> look at the commit log of the commit that added this command:
>
> Add ctrl_iface command for triggering a roam to a specific BSS
>
> 'wpa_cli roam <bssid>' can now be used to test roaming within an ESS
> (e.g., for FT over-the-air). This command will bypass a new scan and
> will select the BSS based on the specified BSSID. It is responsibility
> of the caller to make sure that the target AP is in the BSS table.
> This can be done, e.g., by running a scan before the roam command,
> if needed.
>
>
> (and yes, that BSS table there is referring to cfg80211 BSS table as
> well, i.e., you may need to trigger a scan just before using ROAM; and
> obviously, do not call ROAM if you do not want to "roam" even back to
> the current AP :-)
Yep, I figured out the need for scanning quickly enough, and it does seem
to be working as desired.
Here are some colorful graphs based on results so far. I need
to tweak the reporting logic a bit more, but it's somewhat functional
now:
http://www.candelatech.com/examples/wifi-migration-report/
We have our radius server behind a network emulator, so if .11r fails to
work properly, our roam times are 600+ms or so.
The device under test is two lower-end cisco APs and a very affordable
AP controller, configured for .11r (FT-EAP).
Each station is running a VOIP call so we get some useful packet
drop stats.
Thanks,
Ben
--
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc http://www.candelatech.com
More information about the Hostap
mailing list