Re: [PATCH 4/5] rt2x00: Convert rt2x00 to read control info from skb->cb

[Ivo van Doorn <ivdoorn@gmail.com>]

On Sunday 11 May 2008, Johannes Berg wrote:
> On Sun, 2008-05-11 at 13:33 +0200, Ivo van Doorn wrote:
> > On Sunday 11 May 2008, Ivo van Doorn wrote:
> > > On Saturday 10 May 2008, Johannes Berg wrote:
> > > > 
> > > > > +	/*
> > > > > +	 * Don't automatically kick the beacon queue.
> > > > > +	 */
> > > > > +	tx_info->flags &= ~IEEE80211_TX_CTL_READY_FOR_TX;
> > > > >  
> > > > >  	/*
> > > > >  	 * Fill in skb descriptor
> > > > >  	 */
> > > > >  	skbdesc = get_skb_frame_desc(skb);
> > > > >  	memset(skbdesc, 0, sizeof(*skbdesc));
> > > > > -	skbdesc->flags |= FRAME_DESC_DRIVER_GENERATED;
> > > > 
> > > > Could you live w/o that change? I'm pondering removing that flag again
> > > > since it doesn't have any meaning any more with the fragmentation/busy
> > > > reschedule rewrite I'll do.
> > > 
> > > Not a problem, I had already removed a few instances where I used that,
> > > and will remove the remaining bits as well.
> > 
> > Hmm, things could be much easier for rt2x00 if there was a mac80211 flag
> > that would indicate if the tx_status should be called or not upon txdone.
> > 
> > I used the IEEE80211_TX_CTL_READY_FOR_TX to make sure RTS and CTS frames
> > were not reported to mac80211 when they were send, but would it perhaps be
> > better if I used the flag IEEE80211_TX_CTL_REQ_TX_STATUS for that?
> > I remember that some time ago drivers weren't supposed to check for that
> > flag, or at least not to determine if the tx_status should be called, but has
> > this changed and can the flag be used or should rt2x00 keep track of what
> > should be reported and what not?
> 
> Yes, it probably would be better, but we haven't figured out useful
> settings for that flag yet. I'm fine with unconditionally adding it to
> all frames for now though.

Excellent, I'll prepare a patch that will always set the
IEEE80211_TX_CTL_REQ_TX_STATUS flag in mac80211 then. :)

Ivo