Re: [PATCH v3 1/3] nl80211: Add probe response offload attribute

[Johannes Berg <johannes@sipsolutions.net>]

On Tue, 2011-11-08 at 17:53 +0200, Arik Nemtsov wrote:
> On Tue, Nov 8, 2011 at 17:31, Johannes Berg <johannes@sipsolutions.net> wrote:
> > On Tue, 2011-11-08 at 16:35 +0200, Arik Nemtsov wrote:
> >> Notify user-space about probe-response offloading support in the driver.
> >>
> >> A wiphy flag is used to indicate support and a bitmap of protocols
> >> determines which protocols are supported.
> >
> >> + * @NL80211_ATTR_PROBE_RESP_OFFLOAD_SUPPORT: Indicates support for probe
> >> + *   response offloading by the driver/firmware.
> >> + *   In addition this attribute holds a bitmap of the supported protocols
> >> + *   for offloading using &enum nl80211_probe_resp_offload_support_attr.
> >
> > I wonder if this should make it more clear that probe requests will be
> > replied to, and this is not optional?
> 
> Actually wl12xx "supports" the legacy way as well. We can operate
> without wpa_s explicitly setting the probe-resp (even though we just
> piece it together in other ways).
> In this sense, it's not a mandatory feature. If wpa_s fails to
> generate it for some reason, we don't want it to fail the AP load. The
> AP will operate just fine.

It'll operate, not sure about "just fine" :-) The way I see it that's
mostly a backward compatibility feature, no?

> >> +++ b/include/net/cfg80211.h
> >> @@ -1693,6 +1693,7 @@ enum wiphy_flags {
> >>       WIPHY_FLAG_AP_UAPSD                     = BIT(14),
> >>       WIPHY_FLAG_SUPPORTS_TDLS                = BIT(15),
> >>       WIPHY_FLAG_TDLS_EXTERNAL_SETUP          = BIT(16),
> >> +     WIPHY_FLAG_SUPPORT_PROBE_RESP_OFFLOAD   = BIT(17),
> >
> > Ditto here, maybe WIPHY_FLAG_HAS_PROBE_RESP_OFFLOAD?
> >
> > As we discussed, ath6kl & similar full-mac drivers with AP SME in the
> > device should set this to advertise the probe protocol feature set they
> > support (by passing up), and even for wl12xx it's not optional.
> 
> All this has to do with user-space semantics. Currently hostap always
> sets the probe-resp IEs, without checking offloading support.

Which is fine too, but I think ath6kl should set this value right?

johannes