[Patch v3] supplicant: add dbus getter method for nl80211 iftype
Mon Dec 22 02:59:34 PST 2014
On Mon, 2014-12-22 at 12:42 +0200, Jouni Malinen wrote:
> > So when wiphy capabilities indicate support for AP interface, it does not
> > mean we should expect very first interface to be of AP type.
> > Driver can have multiple virtual interfaces and one of them would support
> > AP operation. So having right to know interface type is something that is
> > not expected??
> That's a kernel side question (i.e., not really a topic for the hostap
> mailing list, but linux-wireless). I agree that vendors can have
> different designs in how what kind of combinations are supported.
> However, you will need to make sure cfg80211 can advertise the
> capabilities properly for your driver. It sounds like that is not the
> case here now [...]
This is indeed the case. The question is, however, less around the
actual combination capabilities and more around how those are
interpreted with existing interfaces. I'm arguing that the current
advertisement of capabilities and combinations makes no representations
about which interfaces can change types at what time and assumes that
any interface belonging to a given hardware can always change to any
other type as long as the combinations are respected.
I would *very strongly* recommend that the driver be fixed to remove
this limitation, it clearly must be possible to fix this in the driver,
at least by decoupling the "netdev" and "internal device/firmware
virtual device" concepts.
If this is not possible then the current interface combinations
advertisement that cfg80211 exposes cannot be used, and a new API must
be added that takes such restrictions into account. I'd hate to see
this, but if I've seen reasonable effort of (failing at) fixing the
driver I might be willing to accept such an API change into
More information about the Hostap