Re: [RFC v3 6/8] wifi: mac80211: extend ifcomb check functions for multi-radio

[Felix Fietkau <nbd@xxxxxxxx>]

On 07.06.24 12:53, Karthikeyan Periyasamy wrote:
> On 6/7/2024 3:52 PM, Felix Fietkau wrote:
> > On 07.06.24 12:07, Karthikeyan Periyasamy wrote:
> > > On 6/6/2024 11:37 PM, Felix Fietkau wrote:
> > > > Add support for counting global and per-radio max/current number of
> > > > channels, as well as checking radio-specific interface combinations.
> > > >
> > > > Signed-off-by: Felix Fietkau <nbd@xxxxxxxx>
> > > > ---
> > > >   net/mac80211/cfg.c         |   7 +-
> > > >   net/mac80211/chan.c        |  17 +++--
> > > >   net/mac80211/ibss.c        |   2 +-
> > > >   net/mac80211/ieee80211_i.h |   6 +-
> > > >   net/mac80211/iface.c       |   2 +-
> > > >   net/mac80211/util.c        | 131 
> > > > +++++++++++++++++++++++++++-----------
> > > >   6 files changed, 116 insertions(+), 49 deletions(-)
> > > >
> > > > diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c
> > > > index 62119e957cd8..950b7b72f0b8 100644
> > > > --- a/net/mac80211/cfg.c
> > > > +++ b/net/mac80211/cfg.c
> > > > @@ -263,7 +263,7 @@ static int ieee80211_start_p2p_device(struct 
> > > > wiphy *wiphy,
> > > >       lockdep_assert_wiphy(sdata->local->hw.wiphy);
> > > > -    ret = ieee80211_check_combinations(sdata, NULL, 0, 0);
> > > > +    ret = ieee80211_check_combinations(sdata, NULL, 0, 0, -1);
> > > >       if (ret < 0)
> > > >           return ret;
> > > > @@ -285,7 +285,7 @@ static int ieee80211_start_nan(struct wiphy *wiphy,
> > > >       lockdep_assert_wiphy(sdata->local->hw.wiphy);
> > > > -    ret = ieee80211_check_combinations(sdata, NULL, 0, 0);
> > > > +    ret = ieee80211_check_combinations(sdata, NULL, 0, 0, -1);
> > > >       if (ret < 0)
> > > >           return ret;
> > > > @@ -4001,7 +4001,7 @@ __ieee80211_channel_switch(struct wiphy *wiphy, 
> > > > struct net_device *dev,
> > > >           goto out;
> > > >       /* if reservation is invalid then this will fail */
> > > > -    err = ieee80211_check_combinations(sdata, NULL, chanctx->mode, 0);
> > > > +    err = ieee80211_check_combinations(sdata, NULL, chanctx->mode, 
> > > > 0, -1);
> > >
> > > Once we reach the global limit, all the -1 passing caller get fail
> > > becuase the iface check param (existing and new) is validated against
> > > the global limit. since global limt as a sort of union of all per-radio
> > > limits.
> > >
> > > Ex:
> > > Global iface = 6 (Radio iface 2GHz:4, 5GHz:4, 6GHz:6)
> > >
> > >
> > > So far 6 iface created (Radio iface 2GHz:2, 5GHz:3, 6GHz:1)
> > >
> > > In this case, new iface creation get fail because caller uses
> > > ieee80211_check_combinations() with -1 as radio idx, so it checked
> > > against global limit
> >
> > Use the sum of the number of interfaces from each radio instead of the 
> > maximum.
>
> Oh, then legacy user have misconception of the global interfaces
> advertised and try to fail for the allowed limits.

Sure, but that might be an issue either way until user space is updated 
and users start looking at the per-radio ifcomb data.
The global data is simply not enough to describe the details of the 
radio split.

- Felix