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;
Subject: Re: [PATCH v2 1/4] mac80211: call ieee80211_tx_h_rate_ctrl() when
 dequeue
From:   Felix Fietkau <nbd@nbd.name>
To:     Ryder Lee <ryder.lee@mediatek.com>,
        Johannes Berg <johannes.berg@intel.com>
Cc:     Lorenzo Bianconi <lorenzo.bianconi@redhat.com>,
        Shayne Chen <shayne.chen@mediatek.com>,
        Evelyn Tsai <evelyn.tsai@mediatek.com>,
        Bo Jiao <bo.jiao@mediatek.com>,
        Sujuan Chen <sujuan.chen@mediatek.com>,
        linux-wireless@vger.kernel.org, linux-mediatek@lists.infradead.org
References: <2176023d8f13d82d093452e1c105609396c30622.1622164961.git.ryder.lee@mediatek.com>
 <a89137bf-401b-c9d0-8848-d79ea9e2a9c3@nbd.name>
Message-ID: <4d652f4a-c258-f248-911f-28238531ae4d@nbd.name>
Date:   Thu, 17 Jun 2021 19:59:26 +0200
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:78.0)
 Gecko/20100101 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <a89137bf-401b-c9d0-8848-d79ea9e2a9c3@nbd.name>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2021-06-17 12:50, Felix Fietkau wrote:
> 
> On 2021-05-28 08:05, Ryder Lee wrote:
>> Make ieee80211_tx_h_rate_ctrl() get called on dequeue to improve
>> performance since it reduces the turnaround time for rate control.
>> 
>> Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
>> ---
>> change since v2 - roll ieee80211_tx_h_rate_ctrl checks into one condition
> There were some OpenWrt crash reported which appear to be related to
> this patch. I was able to reproduce a deadlock with ath9k, and I'm
> currently looking into it.
Some more information about the crash:
- ath9k calls ieee80211_tx_dequeue with the tx queue lock held
- ieee80211_tx_dequeue calls minstrel get_rate
- get_rate calls minstrel_aggr_check
- minstrel_aggr check calls ieee80211_start_tx_ba_session
- ieee80211_start_tx_ba_session tries to send a frame
- ath9k tries to acquire the tx lock it already holds

I've fixed this in v3 by moving the logic of minstrel_aggr_check into
mac80211

- Felix