On 2019-09-19 20:22, Toke Høiland-Jørgensen wrote:
From: Toke Høiland-Jørgensen <toke@xxxxxxxxxx>
Some devices have deep buffers in firmware and/or hardware which
prevents
the FQ structure in mac80211 from effectively limiting bufferbloat on
the
link. For Ethernet devices we have BQL to limit the lower-level queues,
but
this cannot be applied to mac80211 because transmit rates can vary
wildly
between packets depending on which station we are transmitting it to.
To overcome this, we can use airtime-based queue limiting (AQL), where
we
estimate the transmission time for each packet before dequeueing it,
and
use that to limit the amount of data in-flight to the hardware. This
idea
was originally implemented as part of the out-of-tree airtime fairness
patch to ath10k[0] in chromiumos.
This patch ports that idea over to mac80211. The basic idea is simple
enough: Whenever we dequeue a packet from the TXQs and send it to the
driver, we estimate its airtime usage, based on the last recorded TX
rate
of the station that packet is destined for. We keep a running per-AC
total
of airtime queued for the whole device, and when that total climbs
above 8
ms' worth of data (corresponding to two maximum-sized aggregates), we
simply throttle the queues until it drops down again.
The estimated airtime for each skb is stored in the tx_info, so we can
subtract the same amount from the running total when the skb is freed
or
recycled. The throttling mechanism relies on this accounting to be
accurate (i.e., that we are not freeing skbs without subtracting any
airtime they were accounted for), so we put the subtraction into
ieee80211_report_used_skb().
This patch does *not* include any mechanism to wake a throttled TXQ
again,
on the assumption that this will happen anyway as a side effect of
whatever
freed the skb (most commonly a TX completion).
The throttling mechanism only kicks in if the queued airtime total goes
above the limit. Since mac80211 calculates the time based on the
reported
last_tx_time from the driver, the whole throttling mechanism only kicks
in
for drivers that actually report this value. With the exception of
multicast, where we always calculate an estimated tx time on the
assumption
that multicast is transmitted at the lowest (6 Mbps) rate.
The throttling added in this patch is in addition to any throttling
already
performed by the airtime fairness mechanism, and in principle the two
mechanisms are orthogonal (and currently also uses two different
sources of
airtime). In the future, we could amend this, using the airtime
estimates
calculated by this mechanism as a fallback input to the airtime
fairness
scheduler, to enable airtime fairness even on drivers that don't have a
hardware source of airtime usage for each station.
So if it is going to work together with virtual time based mechanism in
the future, the Tx criteria will be met both of below conditions,
1. Lower than g_vt
2. Lower than IEEE80211_AIRTIME_QUEUE_LIMIT
Are we going to maintain two kinds of airtime that one is from
estimation and the other is basically from FW reporting?
Meanwhile, airtime_queued will also limit the situation that we only
have a station for transmission. Not sure if the peak throughput will be
impacted. I believe it may work fine with 11ac in chromiumos, how about
11n and 11a?
Anyway, I think this approach will help to improve performance of the
virtual time based mechanism since it makes packets buffered in host
instead of FW's deep queue. We have observed 2-clients case with
different ratio in TCP fails to maintain the ratio because the packets
arriving at host get pushed to FW immediately and thus the airtime
weight sum is 0 in most of time meaning no TXQ in the rbtree. For UDP,
if we pump load more than the PHY rate, the ratio can be maintained
beacuse the FW queue is full and packtes begin to be buffered in host
making TXQs staying on the rbtree for most of time. However, TCP has its
own flow control that we can not push enough load like UDP.
out:
@@ -3676,6 +3706,9 @@ struct ieee80211_txq *ieee80211_next_txq(struct
ieee80211_hw *hw, u8 ac)
spin_lock_bh(&local->active_txq_lock[ac]);
+ if (local->airtime_queued[ac] > IEEE80211_AIRTIME_QUEUE_LIMIT)
+ goto out;
+
begin:
txqi = list_first_entry_or_null(&local->active_txqs[ac],
struct txq_info,
@@ -3753,6 +3786,9 @@ bool ieee80211_txq_may_transmit(struct
ieee80211_hw *hw,
spin_lock_bh(&local->active_txq_lock[ac]);
+ if (local->airtime_queued[ac] > IEEE80211_AIRTIME_QUEUE_LIMIT)
+ goto out;
+
if (!txqi->txq.sta)
goto out;
--
Yibo
