Error while enabling AP using wpa_supplicant

[Jose Blanquicet]

Hi Jouni,

On Sun, Jan 29, 2017 at 3:27 PM, Jouni Malinen wrote:
> Which driver are you testing with? This works fine in mac80211_hwsim
> test cases. Can you provide a debug log showing the full initialization
> with and without this commit included?

We are using Marvell driver sd8887. The driver exposes three
interfaces mlan0, uap0 and wfd0.
I am sending you the logs with and without commit.

> Could you please clarify what you mean with premature in this context?
> The nl80211_get_wiphy_data_ap() call was moved to a later point in the
> process, not earlier, i.e., it used to be immediately after the
> management frame registration steps while it is now after that and
> channel configuration. Similarly, the nl80211_register_beacons() call
> and NL80211_CMD_REGISTER_BEACONS command gets pushed out to happen after
> channel configuration. That is the operation that your log seemed to
> indicate as failing with EOPNOTSUPP.

Looking at the source code I wrongly understood that
nl80211_get_wiphy_data_ap() call was moved to a earlier point. That's
what I was trying to say with "premature call" but I was wrong. Forget
this part.

> As far as cfg80211 is concerned, nl80211_register_beacons() returns
> -EOPNOTSUPP only if the wiphy does not support WIPHY_FLAG_REPORTS_OBSS.
> I don't see how that could differ based on whether this operation
> happens before or after channel configuration.

On the other hand, yes, NL80211_CMD_REGISTER_BEACONS operation is
indeed where the failure is happening with EOPNOTSUPP error. Something
changed with this commit and is causing such operation to fail and I
have not been able to understand what it is.

Regards,

Jose Blanquicet