[PATCH] Revert "nl80211: Use global netlink rtm event object"

[Ben Greear]

On 10/28/2011 11:14 AM, Jouni Malinen wrote:
> On Fri, Oct 28, 2011 at 09:55:22AM -0700, Ben Greear wrote:
>>> To be more exact, this breaks AP (including P2P GO) mode operations
>>> with wpa_supplicant because of EVENT_INTERFACE_DISABLED event getting
>>> generated at a particularly inconvenient time. Currently, that makes
>>> wpa_supplicant start scanning to find a station mode connection when the
>>> interface "comes back" (i.e., immediately after that on the next netlink
>>> event). This is obviously wrong behavior and should be fixed somehow
>>> regardless.
>
>> I don't see how that patch would change any behaviour like this,
>> but I'll work on reproducing the problem and see what I can learn.
>
> Agreed, the netlink related patch is not really broken, but it is
> highlighting another issue that has been there with wpa_supplicant AP
> mode from the beginning. It was just not showing up frequently enough to
> get enough energy to fixing it. Your patch made it happen on every
> P2P GO start, so this kind of changes ;-).
>
> Instead of reverting the patch, I added yet another workaround for the
> netlink download processing: commit
> 59d249255c5ea484ac2050f19ecbe84a7d9323f0.

Thanks.  I sync'd this down and verified it also fixes the problem
for me (I was able to reproduce it).

I also tested a small number of regular virtual stations and that
all seems good too.

Hopefully there is nothing that does a quick interface bounce and
actually wants the down/up state change logic to happen in hostap,
because if so, that work-around will probably break it.

Thanks,
Ben

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