On Wed, 2020-05-27 at 16:41 +0200, Johannes Berg wrote: > > Hmm. Yes, we just had > > + case IEEE80211_HE_6GHZ_OPER_CTRL_CHANWIDTH_160MHZ: > + if (abs(he_6ghz_oper->ccfs1 - he_6ghz_oper->ccfs0) == 8) > + he_chandef.width = NL80211_CHAN_WIDTH_160; > + else > + he_chandef.width = NL80211_CHAN_WIDTH_80P80; > + break; > > > but that breaks if you don't support 80+80 or 160. > > OTOH, we check this later again, I think, and downgrade if we don't > support it, so no harm done? > > I think I'd prefer the parsing to be exact, and then downgrade as > necessary. That makes things a bit simpler. Except that won't work for mesh. I actually kinda like this better than what I did, because what I did required all kinds of contortions with DISABLE_HT/VHT/HE ... Still checking. johannes
