Re: [PATCH] mac80211: do not iterate active interfaces when in re-configure

[Johannes Berg <johannes@sipsolutions.net>]

On Mon, 2020-05-25 at 09:53 -0700, greearb@candelatech.com wrote:
> From: Ben Greear <greearb@candelatech.com>
> 
> This appears to fix a problem where ath10k firmware would crash,


"appears to", heh

Really though, in general, you need to start thinking about mac80211 and
the drivers as two separate things. You've also submitted another patch
where you say "this iwlwifi thing requires mac80211 to change", and here
you're submitting a patch saying "this ath10k thing requires mac80211 to
change", but I don't see you considering much the fact that mac80211 is
actually used for both.

It'd be good to have a discussion of such things in the commit log for
changes that will affect multiple drivers, rather than focusing on a
single bug for a single driver. In general, not just in this patch.

> diff --git a/net/mac80211/util.c b/net/mac80211/util.c
> index 5db2cd0..186a696 100644
> --- a/net/mac80211/util.c
> +++ b/net/mac80211/util.c
> @@ -831,7 +831,7 @@ static void __iterate_interfaces(struct ieee80211_local *local,
>  			break;
>  		}
>  		if (!(iter_flags & IEEE80211_IFACE_ITER_RESUME_ALL) &&
> -		    active_only && !(sdata->flags & IEEE80211_SDATA_IN_DRIVER))
> +		    (active_only && (local->in_reconfig || !(sdata->flags & IEEE80211_SDATA_IN_DRIVER))))
>  			continue;

Anyway, this seems wrong to me. If anything, it should skip those
interfaces that weren't re-added yet, but not all of them. I'm pretty
sure this would cause iwlwifi to misbehave with multiple interfaces, as
it sometimes relies on iterating to understand what else is going on
before configuring something.

It might even be that this can only be done subject to driver choice.

johannes