hostapd error

[Ben Greear]

On 03/26/2016 02:09 AM, Jouni Malinen wrote:
> On Fri, Mar 25, 2016 at 05:49:57PM -0700, Ben Greear wrote:
>> On 03/25/2016 05:24 PM, Ann Lo wrote:
>>> Thanks for your information and suggestion. Using -dd option, there
>>> are more information from hostapd, including:
>>>
>>> nl80211: Beacon set failed: -16 (Device or resource busy)
>>>
>>> Further investigation result is that this happens in the following scenario:
>>> 1) hostapd runs OK at channel 36 using 80 MHz width.
>>> 2) wpa_supplicant (using the same radio) tries to connect to a remote
>>> AP on channel 36 using 20 MHz width.
>>> 3) hostapd is restarted at this point.
>>> 4) When hostapd comes up again, it gets the above error.
>>> 5) At this point, wpa_supplicant connects successfully.
>>>
>>> The radio requires AP and client using the same channel. Does this
>>> mean that the channel width must be the same?
>>
>> Channel width does not need to be the same, but the primary channel
>> must be.
>>
>> You probably need a patch I put into hostapd to force not switching
>> the primary channel when running multiple interfaces on a single radio.
>>
>> https://github.com/greearb/hostap-ct/commit/cd25b893c97da95b5b7cda0e61ea7bcf5355bc16
>
> Instead of that configuration parameter that can be used to disable
> mandatory 40 MHz co-existence requirements, shouldn't this be handled
> automatically by detecting that another local vif of the same radio
> prevents the switch and based on that, prevent the switch? I'd be fine
> doing this based on the local concurrent use constraints, but it would
> be difficult to justify parameters that can be misused.

That would not fix the case where you want to:

create 80Mhz AP on CH 36  #  Might want to change primary channel
create 20Mhz AP on CH 36  #  Cannot change primary channel.

When 80Mhz AP is created, no other vif exists, so it has no way to know
it cannot change primary channel.

Thanks,
Ben

-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com