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: <d7ce90d8-552a-2cab-6310-3a84584172a1@green-communications.fr>
Date:   Tue, 18 Jan 2022 10:08:15 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.4.0
Content-Language: en-US
From:   Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Subject: Issues with mt7915 thermal throttling
To:     linux-wireless@vger.kernel.org
Cc:     Ryder Lee <ryder.lee@mediatek.com>,
        Lorenzo Bianconi <lorenzo.bianconi83@gmail.com>,
        Felix Fietkau <nbd@nbd.name>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

I noticed some strange issues with mt7915's thermal throttling, 
particularly the cooling_device.

First, it seems that the thermal subsystem expect that higher 
cooling_device states provide more cooling, but mt7915e apparently does 
the opposite and use it as a duty cycle, where state=1 does severe 
throttling/max cooling (iperf throughput basically go down to near zero) 
and state=100 is full power.

Also, state=0, from the comments, apparently disable thermal management, 
except that in practice, it does not change the throttle state, since 
throughput stays low when switching from state=1 to state=0, and stays 
high when switching from 100 to 0.

As a result, as soon as the default thermal zone runs a little hot, the 
performance of mt7915e is destroyed and does not recover much when the 
temperature drops down.

I can come up with a patch to fix the first issue, but not the state=0 
one, and i would like some pointers/confirmation.