Re: [PATCH 2/2] HE: propagate BSS color settings to nl80211

On Sun, Feb 16, 2020 at 01:16:41PM +0200, Jouni Malinen wrote:
> On Tue, Feb 04, 2020 at 09:04:55AM +0100, John Crispin wrote:
> > This patch adds the code required to send the BSS color settings to the kernel.

> > diff --git a/src/drivers/driver_nl80211.c b/src/drivers/driver_nl80211.c
> > @@ -4108,6 +4108,9 @@ static int wpa_driver_nl80211_set_ap(void *priv,
> > +	bss_color = nla_nest_start(msg, NL80211_ATTR_HE_BSS_COLOR);
> 
> This breaks all AP functionality with commonly used libnl versions if
> the kernel includes a check for NLA_F_NESTED. And this happens even if
> no IEEE 802.11ax is used.
> 
> nl80211: kernel reports: NLA_F_NESTED is missing
> nl80211: Beacon set failed: -22 (Invalid argument)

> At minimum, this addition of NL80211_ATTR_HE_BSS_COLOR would need to
> depend on HE being enabled in the configuration, but even that does not
> sound nice. Maybe doing an ugly hack like this would be justifiable
> here:
> 
> bss_color = nla_nest_start(msg, NL80211_ATTR_HE_BSS_COLOR | NLA_F_NESTED);
> 
> This could be conditional based on the used libnl version. Maybe
> something with NL_CAPABILITY_VERSION_3_5_0 (which is defined in
> netlink/utils.h) or nl_has_capability() at runtime.

I added a generic libnl workaround for nla_nest_start(), so this issue
went away and I applied this patch with some cleanup and a fix
(nla_nest_start() return value needs to be checked).
 
-- 
Jouni Malinen                                            PGP id EFC895FA