Received-SPF: pass (google.com: best guess record for domain of linux-wireless-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Toke =?utf-8?Q?H=C3=B8iland-J=C3=B8rgensen?= <toke@redhat.com>
To:     Kan Yan <kyan@google.com>
Cc:     Yibo Zhao <yiboz@codeaurora.org>, Felix Fietkau <nbd@nbd.name>,
        Johannes Berg <johannes@sipsolutions.net>,
        linux-wireless@vger.kernel.org,
        make-wifi-fast@lists.bufferbloat.net,
        John Crispin <john@phrozen.org>,
        Lorenzo Bianconi <lorenzo@kernel.org>,
        linux-wireless-owner@vger.kernel.org
Subject: Re: [PATCH RFC/RFT 4/4] mac80211: Apply Airtime-based Queue Limit (AQL) on packet dequeue
In-Reply-To: <CA+iem5tDU9RYdUJ5y1E3E1iFRsSX_rA8dbDjaviGBxKyMS_3kA@mail.gmail.com>
References: <156889576422.191202.5906619710809654631.stgit@alrua-x1> <156889576869.191202.510507546538322707.stgit@alrua-x1> <08f0ed6e-b746-9689-6dc8-7c0ea705666d@nbd.name> <87wodv19jl.fsf@toke.dk> <b0239b81-e3db-1f10-22cb-ad06536b97a8@nbd.name> <87tv8z13wv.fsf@toke.dk> <CA+iem5uir8GOq2psAVuUParY3MimsGhvdefUopiwHov128CQQA@mail.gmail.com> <87r2421d4f.fsf@toke.dk> <ed46699c32cd3d8d8bcc9bce1375ba04@codeaurora.org> <CA+iem5sGkRQy4B=Mz-C90sfqgFmuywa=W7Yjy2CDd8hfppQkvQ@mail.gmail.com> <87muemykqn.fsf@toke.dk> <CA+iem5tDU9RYdUJ5y1E3E1iFRsSX_rA8dbDjaviGBxKyMS_3kA@mail.gmail.com>
Date:   Tue, 01 Oct 2019 08:57:53 +0200
Message-ID: <87imp9vtou.fsf@toke.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=WINDOWS-1252
Content-Transfer-Encoding: quoted-printable
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk

Kan Yan <kyan@google.com> writes:

>> I guess the risk is lower when with a 24ms per-iface limit; but with
>> enough stations I guess it could still happen, no? So we should probably
>> handle this case...
> Each txq (per sta, per tid) is allowed to release at least the lower
> AQL limit amount of packet (default 4ms), which is not affected by
> other station's PS behavior and 4ms should be sufficient for most use
> cases.

Ah, I thought you'd meant each station can queue MIN(4ms, 24ms-<other
stations>). I see that is not the case; it's up to 10ms as long as the
total is less than 20ms, and up to 4ms otherwise.=20

> The 24ms per-interface limit is an optimization to get good benchmark
> score in peak performance test, which usually only involve 1-2
> stations.

Gotta get those benchmark numbers in ;)

> The higher limit probably won't matter anymore when there are many
> stations. I haven't noticed side effects due to PS behavior in the
> ChromiumOS version. Still, it is better to be able to take frames in
> PS queue in to account,

As long as one station always gets its 4ms, I'm not too worried about
PS; but that was not the case in my patch :)

>> Cool. Are you going to submit a ported version of your implementation?
>> Then we can work from the two submissions and see if we can't converge
>> on something...
> Working on porting, should have something ready before the end of this
> week.

Great!

-Toke