Re: [RFC] cfg80211: Fix incompatible interfaces combination

[Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>]

On Monday 04 June 2012 09:22 PM, Johannes Berg wrote:
> On Mon, 2012-06-04 at 21:17 +0530, Mohammed Shafi Shajakhan wrote:
>> From: Mohammed Shafi Shajakhan<mohammed@qca.qualcomm.com>
>>
>> *if any interface type is not advertised by the driver
>> via ieee80211_iface_combination make sure we will have it
>> as a single interface only. lets that we will not add
>> an incompatible interface if some other interface is
>> already present. we cannot add any other interface,
>> if the already present interface is an incompatible
>> interface. for example in ath9k we don't advertise ADHOC in
>> ieee80211_iface_combination structure in the driver,
>> so it can only exist as an single interface
>
> I'm not sure I understand what you're changing here?

if an interface say ad-hoc is not advertised in iface combination,
with the current still we can add that interface with some other 
interface(removing the driver checks).

in ath9k we have

static const struct ieee80211_iface_limit if_limits[] = {
         { .max = 2048,  .types = BIT(NL80211_IFTYPE_STATION) |
                                  BIT(NL80211_IFTYPE_P2P_CLIENT) |
                                  BIT(NL80211_IFTYPE_WDS) },
         { .max = 8,     .types =
#ifdef CONFIG_MAC80211_MESH
                                  BIT(NL80211_IFTYPE_MESH_POINT) |
#endif
                                  BIT(NL80211_IFTYPE_AP) |
                                  BIT(NL80211_IFTYPE_P2P_GO) },
};

i am able to create an IBSS interface even if a STA interface is up and 
running in ath9k(removing the driver checks).

this patch ensures that if we have an incompatible interface, it is the 
only one can that can exist.

>
>> *if the driver is not advertising interface combination just
>> return so that we don't break their multivif operation.
>
> Actually, we want to break multivif for drivers that don't advertise it.

oh, ok.

>
> johannes
>


-- 
thanks,
shafi