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;
To:     Georgi Valkov <gvalkov@abv.bg>
Cc:     linux-wireless@vger.kernel.org, johannes@sipsolutions.net,
        openwrt-devel@lists.openwrt.org
References: <20200921154545.91971-1-nbd@nbd.name>
 <7A099B27-5278-4D2A-91A1-6B795CB0E221@abv.bg>
From:   Felix Fietkau <nbd@nbd.name>
Subject: Re: [PATCH] mac80211: fix regression in sta connection monitor
Message-ID: <099b1679-87c4-1ac5-978b-a1110db93f54@nbd.name>
Date:   Tue, 22 Sep 2020 10:56:13 +0200
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:78.0)
 Gecko/20100101 Thunderbird/78.2.2
MIME-Version: 1.0
In-Reply-To: <7A099B27-5278-4D2A-91A1-6B795CB0E221@abv.bg>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 2020-09-22 09:28, Georgi Valkov wrote:
> Hi Felix!
> 
> In the original code before the regression, as well as in the fix I
> proposed, probe_send_count was always reset if ack is true, while in
> your fix it will not be reset when non-null frames are received. The
> connection failed again after 10 hours and the STA remained offline for
> two minutes instead of two seconds. I should also note that I spent
> multiple days testing the fix I proposed and I had no disconnects.
> Please revise your fix again!
Thanks for the info. Handling non-null frames in that function is
incomplete, because it does not cover the 802.3 tx status path.
I'll make v2 patch that checks the timestamp of the last ACK in
ieee80211_sta_work instead.

- Felix