[PATCH] Fix scan notification in case of canceling scan

[Jouni Malinen]

On Mon, Apr 11, 2011 at 03:50:39PM -0700, Dmitry Shmidt wrote:
> It is just a good function that is called in all possible "Associate" places.
> Also this change is harmful - in case of "no scan in progress" nothing
> will happen.

I'd assume you meant "harmless" here. I'm not sure I would agree with
that, though, in general. There could be cases where a pending scan
request gets is canceled while a real scan is going through. I'm not
saying that the current implementation would be doing that, but I do not
really want to leave this type of traps in the code for future
extensions.

I would prefer to see a more robust way of doing this by the driver_*.c
code indicated canceled scans. With nl80211, that is already done. If
the driver does not indicate a canceled scan on the association command,
driver_*.c should be able to do that internally by tracking whether
there is a pending scan and if that happens to be the case when
associate() gets called, indicate the canceled scan with
EVENT_SCAN_RESULTS that has data->scan_info.aborted set to 1.

-- 
Jouni Malinen                                            PGP id EFC895FA