Re: [PATCH] netdev: handle disconnect event during the 4-way handshake

[James Prestwood <prestwoj@gmail.com>]

Hi Alvin,

On 7/17/24 12:35 PM, Alvin Šipraga wrote:
> Hi James,
>
> On Wed, Jul 17, 2024 at 08:02:17AM GMT, James Prestwood wrote:
>> + CC Alvin
>>
>> You had originally reported [1] a similar issue but it was unclear exactly
>> what was happening at the time. I believe this is the same problem. This may
>> not be the final version of a fix, but I wanted to CC you in case you still
>> were seeing this and wanted to test it.
>>
>> [1] https://lore.kernel.org/iwd/20230403141927.235014-1-alvin@pqrs.dk/
> It's been a while since I dealt with this problem, and in the mean time
> we have been using the original patch I sent to the list as seen in your
> link. Unfortunately I don't recall how I reproduced this issue, but
> maybe my colleague Pedro (+cc) remembers.
>
> Regardless, a static code analysis suggests that your patch yields the
> same result, so I think this is good - one less patch for us to apply :)
>
> Do you have some easily communicable steps to reproduce this issue on
> your end?

I was able to write an autotest for it that took that code path. Its 
basically just as you thought from your previous thread. If a deauth 
frame comes in during the 4-way handshake or key settings during a 
reassociate roam it triggers that "Unexpected disconnect" path in 
station, causing IWD to hang up and not clean up its state.

I just wanted to check in with you that its being fixed. Denis and I 
have been discussing it and my ML patch still isn't quite how he wants 
it, but we think we know how it should be handled.

Thanks,
James

>
> Thanks for following up on this!
>
> Kind regards,
> Alvin