Return-path: Received: from mail-wm0-f41.google.com ([74.125.82.41]:35832 "EHLO mail-wm0-f41.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751257AbcBHLR0 convert rfc822-to-8bit (ORCPT ); Mon, 8 Feb 2016 06:17:26 -0500 Received: by mail-wm0-f41.google.com with SMTP id c200so10392144wme.0 for ; Mon, 08 Feb 2016 03:17:25 -0800 (PST) MIME-Version: 1.0 In-Reply-To: References: <1454920723-48071-1-git-send-email-nbd@openwrt.org> Date: Mon, 8 Feb 2016 12:17:24 +0100 Message-ID: (sfid-20160208_121730_589352_D85A7404) Subject: Re: [RFC v4] mac80211: add A-MSDU tx support From: Michal Kazior To: Krishna Chaitanya Cc: Emmanuel Grumbach , Felix Fietkau , linux-wireless , Johannes Berg Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 8 February 2016 at 12:09, Krishna Chaitanya wrote: > On Mon, Feb 8, 2016 at 3:56 PM, Emmanuel Grumbach wrote: >> On Mon, Feb 8, 2016 at 11:54 AM, Krishna Chaitanya >> wrote: >>> On Mon, Feb 8, 2016 at 2:56 PM, Emmanuel Grumbach wrote: >>>> On Mon, Feb 8, 2016 at 10:38 AM, Felix Fietkau wrote: >>>>> Requires software tx queueing support. frag_list support (for zero-copy) >>>>> is optional. >>>>> >>>>> Signed-off-by: Felix Fietkau >>>>> --- >>>> >>>> >>>> Ok - looks fine, but... and here comes the hard stuff. >>>> The frame size in the PLCP is limited in a way that you can't - from a >>>> spec POV - enable A-MSDU for low rates. Of course, you don't want to >>>> do that for low rates at all regardless of the spec. >>>> Since you build the A-MSDU in the mac80211 Tx queue which is not aware >>>> of the link quality, how do we prevent A-MSDU if the rate is low / >>>> dropping. >>>> I'd even argue that when the rates get lower, you'll have more >>>> packets piling up in the software queue and ... even more chance to >>>> get A-MSDU in the exact case where you really want to avoid it? >>> >>> Similar to triggering AMPDU setup, we should put this control >>> in RC (minstrel) to start/stop AMSDU based on link quality/if the rates >>> drop below a pre-defined MCS (or) only for best-throughput rates. >> >> Ok - but the size of A-MPDU is determined in the firmware / hardware >> typically (at least for Intel devices and I have to admit my ignorance >> about other designs...). So that if you get bad link conditions, the >> firmware can play for that size. A-MSDU built in the host is a >> different story. I remember that athXk devices attach the rate on the >> Tx packet itself so that the rate control is serialized with the data >> path. >> iwlwifi works differently: we have a map of rates in the firmware. >> This map is maintained in the driver updated based on the feedback we >> get from the device. When an update is needed, iwlwifi sends a command >> to the firmware and that command bypasses all the Tx packets currently >> queue so that a packet that is already waiting in the queue will be >> sent with the updated map. This allows iwlwifi to react downgrade >> faster when needed. >> So - for athXk (and maybe the MediaTek devices Felix is working on), >> this may not be a big problem: you'd need to add another input to the >> A-MSDU building code coming from the rate control. >> FWIW: in iwlwifi I simply decided to be on the safe side and allowed >> the driver to build A-MSDUs only if the rate is fairly high and we are >> not downgrading. But you saw (and commented on) that patch already I >> think :) > > Yes, this approach is going to be tough for those devices with HW RC. > We might need something like amsdu_action to get some feedback from > device about using of amsdu? The HW/FW rate control still may not have sufficient knobs for driver to implement an amsdu_action(). What about if mac80211 tracked per-station-tid packet sojourn time and based on that derive max A-MSDU bytelimit? This could also be useful for the bufferbloat problem (discussed in the other thread) as well. Michał