Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Sender: Guenter Roeck <groeck7@gmail.com>
To:     Jan Kiszka <jan.kiszka@siemens.com>,
        Wim Van Sebroeck <wim@linux-watchdog.org>,
        linux-watchdog@vger.kernel.org
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Tero Kristo <t-kristo@ti.com>
References: <fe8cf65f-f949-9326-8f32-fda7134c8da6@siemens.com>
 <211cd54b-29b4-e58a-341b-beffc05cfe85@roeck-us.net>
 <def0c2e9-e035-7ffc-3216-27f461555ae5@siemens.com>
From:   Guenter Roeck <linux@roeck-us.net>
Subject: Re: [PATCH] watchdog: Respect handle_boot_enabled when setting last
 last_hw_keepalive
Message-ID: <ccbd43d8-af29-2a5d-991f-def7f7adec40@roeck-us.net>
Date:   Fri, 30 Jul 2021 15:39:01 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <def0c2e9-e035-7ffc-3216-27f461555ae5@siemens.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 7/30/21 2:37 PM, Jan Kiszka wrote:
> On 30.07.21 22:49, Guenter Roeck wrote:
>> On 7/30/21 12:39 PM, Jan Kiszka wrote:
>>> From: Jan Kiszka <jan.kiszka@siemens.com>
>>>
>>> We must not pet a running watchdog when handle_boot_enabled is off
>>> because this requests to only start doing that via userspace, not during
>>> probing.
>>>
>>
>> The scope of the changed function is quite limited. See the
>> definition of watchdog_set_last_hw_keepalive(). On top of that,
>> __watchdog_ping() does a bit more than just ping the watchdog,
>> and it only pings the watchdog in limited circumstances. On top of that,
>> the scope of handle_boot_enabled is different: If enabled, it tells
>> the watchdog core to keep pinging a watchdog until userspace opens
>> the device. This is about continuous pings, not about an initial one.
>> Given that, I'd rather have the watchdog subsystem issue an additional
>> ping than risking a regression.
>>
>> The only driver calling watchdog_set_last_hw_keepalive() is rti_wdt.c.
>> Does this patch solve a specific problem observed with that watchdog ?
> 
> Yes, it unbreaks support for handle_boot_enabled=no by not starting the
> automatic pinging of the kernel until userspace opens the device.
> Without this fix, the core will prematurely start kernel-side pinging,
> and hanging userspace will never be detected.
> 
Good point. You are correct.

I think it should also check for watchdog_hw_running(wdd), though.
The function should not really be called if the watchdog isn't
running, but it should still not ping the watchdog in that case.

Something like

     if (watchdog_hw_running(wdd) && handle_boot_enabled)
	return __watchdog_ping(wdd);

     return 0;

Also, I think it would make sense to add your additional comment
to the patch description. The problem isn't only that the watchdog
is pinged once, the problem is that it starts _automatic_
pinging which it really should not do.

Thanks,
Guenter