On 2020-09-28 05:17, Johannes Berg wrote:

quoted

@@ -176,6 +177,15 @@ ieee80211_determine_chantype(struct 

ieee80211_sub_if_data *sdata,
 	memcpy(&sta_ht_cap, &sband->ht_cap, sizeof(sta_ht_cap));
 	ieee80211_apply_htcap_overrides(sdata, &sta_ht_cap);

+	if (s1g_oper && sband->band == NL80211_BAND_S1GHZ) {
+		ieee80211_chandef_s1g_oper(s1g_oper, chandef);
+		ret = IEEE80211_STA_DISABLE_HT | IEEE80211_STA_DISABLE_40MHZ |
+		      IEEE80211_STA_DISABLE_VHT |
+		      IEEE80211_STA_DISABLE_80P80MHZ |
+		      IEEE80211_STA_DISABLE_160MHZ;
+		goto out;
+	}

I've applied this now (and made some changes in a few patches, please
check), but this particular thing seems a bit odd, in that it looks for
HT/VHT elements even on an S1G channel if the s1g oper element isn't
present? That seems like it shouldn't be the case, can you take a look?

Thanks for the (numerous) fixups.

I don't think applying a (NULL) HT cap is a big deal since the same code 
path
is already executed for eg. non HT, but it does look strange and I'll 
send a
fixup.

-- 
thomas