[PATCH 08/50] AP: Add some basic MLD configuration options

Otcheretianski, Andrei andrei.otcheretianski at intel.com
Sun Feb 19 08:17:50 PST 2023

> > +	} else if (os_strcmp(buf, "mld_ap") == 0) {
> > +		bss->mld_ap = !!atoi(pos);
> > +	} else if (os_strcmp(buf, "mld_id") == 0) {
> a EHT AP is MLD AP, should we still need an extra config, Also why mld_id is
> needed as extra config? Can we identify using interface name that all the bss
> belong to same ML AP with unique interface name?

It's possible to have multiple BSS's on the same interface not in MLD scenarios. And it's possible that the user doesn't want to include all of them by default into the same MLD instance.
So I think mld_ap config is needed.
Regarding the mld_id, I think it's reasonable to assume that all the "mld_ap" BSS's that share a common interface are part of the same MLD instance. In such case indeed it's not needed.

> > +		bss->mld_id = atoi(pos);
> > +	} else if (os_strcmp(buf, "mld_link_id") == 0) {
> Should user care about link id, can we have it generated by hostapd instead
> based on the interfaces list and freq.

I thought about it, and it might be a little bit tricky, so it's just simpler to have it, and add a patch later that will do the auto-assigning.
The problem is that the BSS's can be dynamically added and removed and you really would want to keep the id's between different runs.

> > +		bss->mld_link_id = atoi(pos);
> > +	} else if (os_strcmp(buf, "mld_eml_capa") == 0) {
> The mld_capa and eml capa should come from the driver/hw right?
> > +		bss->mld_eml_capa = atoi(pos);

Right.. In current series we still don't read it from kernel reported capabilities (it just wasn't there yet when we started implementing it).
Now, all these configurations are indeed not really needed. I'll send the patch that reads it from the kernel. Shall we keep mld/eml_capa for testing only?


More information about the Hostap mailing list