Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <e338e57075680c4a9e1323c7f4d6bfff1268d4e2.camel@sipsolutions.net>
Subject: Re: [PATCH] mac80211: don't apply flow control on management frames
From:   Johannes Berg <johannes@sipsolutions.net>
To:     mbizon@freebox.fr, linux-wireless@vger.kernel.org
Date:   Mon, 22 Mar 2021 10:46:28 +0100
In-Reply-To: <5f037c3f8485a8c929548be6a153290ba9ae4f13.camel@freebox.fr>
References: <20210319232800.0e876c800866.Id2b66eb5a17f3869b776c39b5ca713272ea09d5d@changeid>
         <5f037c3f8485a8c929548be6a153290ba9ae4f13.camel@freebox.fr>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.38.4 (3.38.4-1.fc33) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

On Mon, 2021-03-22 at 10:43 +0100, Maxime Bizon wrote:
> > Short-circuit the processing of management frames. To keep the
> > impact minimal, only put them on the frags queue and check the
> > tid == management only for doing that and to skip the airtime
> > fairness checks, if applicable.
> 
> After your patch, what are the actual effects of HW_STA_MMPDU_TXQ and
> HW_BUFF_MMPDU_TXQ ?

Well, my patch doesn't change the effect of those significantly.

The idea for iwlwifi was that it doesn't actually like ->tx() to get
called, but much prefers a TXQ where the frame is, and then it can pull
it whenever it can transmit it.

This was the key requirement here, and it doesn't change: instead of
tx() getting called with the frames, the frames go to the TXQ instead
and wake_tx_queue() is called, and then the driver later pulls the
frames and pushes them to the hardware.

What does change is that management frames are no longer subject to
codel and inter-flow issues, and also note that the hash of a management
frame isn't actually well-defined.

johannes