Re: [PATCH v2] mac80211: re-enable aggregation

["Luis R. Rodriguez" <lrodriguez@atheros.com>]

On Wed, Sep 17, 2008 at 03:57:25PM -0700, Tomas Winkler wrote:
> On Thu, Sep 18, 2008 at 1:45 AM, Luis R. Rodriguez
> <lrodriguez@atheros.com> wrote:
> > On Wed, Sep 17, 2008 at 1:57 PM, Tomas Winkler <tomasw@gmail.com> wrote:
> >> On Tue, Sep 16, 2008 at 9:37 AM, Johannes Berg
> >> <johannes@sipsolutions.net> wrote:
> >>> On Mon, 2008-09-15 at 20:17 -0700, Luis R. Rodriguez wrote:
> >>>
> >>>> +/**
> >>>> + * is_part_ampdu - tells us whether this buffer is part of an AMPDU
> >>>> + *
> >>>> + * @skb: the buffer we want to check
> >>>> + * @hw: the &struct ieee80211_hw to check the queue mapping on
> >>>> + */
> >>>> +static inline bool is_part_ampdu(struct sk_buff *skb, struct ieee80211_hw *hw)
> >>>> +{
> >>>> +     return (skb_get_queue_mapping(skb) >=
> >>>> +             ieee80211_num_regular_queues(hw));
> >>>> +}
> >>>> +
> >>>
> >>> This is making the patch needlessly large and makes it change drivers
> >>> etc, please just keep the flag and set it based on this in master xmit
> >>> after we've cleared the info.
> >>>
> >> Yes, please keep the flag. Otherwise our rate scale algorithm won't work
> >
> > OK -- you can ignore this I like the path you took better.
> 
> What I need to next is to add is rtnl_lock
> qdisc_root_lock requires this. This root lock is held during requeuing
> which can be triggered by removal or addition of an aggregation queue.

I saw that, but rtnl_lock() is mutex basd so no good. Hence we need
something that works around this.

  Luis