On Tue, Dec 19, 2017 at 5:58 PM, Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote: > On Tue, 2017-12-19 at 13:37 +0100, Arend van Spriel wrote: >> On 12/19/2017 12:19 PM, Sergey Matyukevich wrote: >> > > > Not yet. At the moment enum nl80211_ap_sme_features in uapi/linux/nl80211.h >> > > > is commented out. For MAC-based ACL the following things are being checked >> > > > on wiphy registration: complete flag WIPHY_FLAG_HAVE_AP_SME, non-zero >> > > > max_acl_mac_addrs, and set_mac_acl cfg80211 callback. >> > > >> > > I guess that's enough then? Userspace can check max_acl_mac_addrs as >> > > well, so it can just use that? >> > >> > Correct, that is what hostapd is doing. I was simply surprised by the fact >> > that MAC-based ACL support implies full-fledged AP SME support. Though >> > your almost convinced me that this is ok and other wireless cards simply >> > do not exist. >> >> So the question seems to be here: what shall drivers/firmware implement >> to allow flag WIPHY_FLAG_HAVE_AP_SME being set. The kerneldoc is a bit >> short in providing guidance: >> >> * @WIPHY_FLAG_HAVE_AP_SME: device integrates AP SME > > They should implement the AP SME? :) > > That is, handling auth/assoc/etc. So basically everything to setup 802.11 connection. So what about the .*_station() callbacks? Anyway, I can understand that people start looking at the checks done in wiphy_register() as a last resort in finding (some sort of) documentation. > With the SAE-"offload"-to-host those lines are blurring again though. Yeah. Thanks for (inadvertently) reminding me to chime in on that. Regards, Arend
