Re: [PATCH] wifi: mac80211: queue frames while off-channel

[Stanislaw Gruszka <stf_xl@wp.pl>]

On Fri, Sep 04, 2026 at 08:56:13AM +0200, Johannes Berg wrote:
> On Thu, 2026-08-20 at 15:12 +0200, Stanislaw Gruszka wrote:
> > 
> > The change depends on the iTXQ resumption handling introduced by
> > commit 4444bc2116ae ("wifi: mac80211: Proper mark iTXQs for resumption")
> > and the wake_tx_queue support introduced by
> > commit a790cc3a4fad ("wifi: mac80211: add wake_tx_queue callback to drivers"),
> > so use those in 'Fixes' tag.
> 
> That doesn't really make any sense, IMHO.
Going to drop those 'Fixes:' tags and comment. but I don't know
what the tag should be then.
I saw in docs that 'Fixes:' is required for -stable submissions
for net tree, but maybe I could just skip it for mac80211?

> > --- a/net/mac80211/tx.c
> > +++ b/net/mac80211/tx.c
> > @@ -267,20 +267,14 @@ ieee80211_tx_h_dynamic_ps(struct ieee80211_tx_data *tx)
> >  }
> >  
> >  static ieee80211_tx_result debug_noinline
> > -ieee80211_tx_h_check_assoc(struct ieee80211_tx_data *tx)
> > +ieee80211_tx_h_check_offchannel(struct ieee80211_tx_data *tx)
> 
> I (vaguely?) had these ordered by execution order, might make sense to
> move around and try to keep that.
Ok, will move.

> >  {
> > -
> >  	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)tx->skb->data;
> > -	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(tx->skb);
> > -	bool assoc = false;
> > -
> > -	if (unlikely(info->flags & IEEE80211_TX_CTL_INJECTED))
> > -		return TX_CONTINUE;
> 
> 
> Not sure you should remove that, still you want injection to go out?
> 
> Not sure injection should even be queued - maybe you do want to send
> while scanning on a random channel? Or maybe not? Maybe someone depends
> on that behaviour? Or maybe not? Kinda hard to say...
I will keep it v2, change indeed could cause regression for someone.

Regards
Stanislaw