Sender: Guenter Roeck <groeck7@gmail.com>
Subject: Re: [PATCH v2 3/4] dt-bindings: watchdog: da9062: add watchdog
 timeout mode
To:     Andrej Picej <andrej.picej@norik.com>,
        Adam Thomson <Adam.Thomson.Opensource@diasemi.com>
Cc:     Support Opensource <Support.Opensource@diasemi.com>,
        "wim@linux-watchdog.org" <wim@linux-watchdog.org>,
        "linux-watchdog@vger.kernel.org" <linux-watchdog@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "robh+dt@kernel.org" <robh+dt@kernel.org>,
        "devicetree@vger.kernel.org" <devicetree@vger.kernel.org>,
        "shawnguo@kernel.org" <shawnguo@kernel.org>,
        "s.hauer@pengutronix.de" <s.hauer@pengutronix.de>,
        "kernel@pengutronix.de" <kernel@pengutronix.de>,
        "festevam@gmail.com" <festevam@gmail.com>,
        "linux-imx@nxp.com" <linux-imx@nxp.com>,
        "linux-arm-kernel@lists.infradead.org" 
        <linux-arm-kernel@lists.infradead.org>
References: <20211130134242.3516619-1-andrej.picej@norik.com>
 <20211130134242.3516619-3-andrej.picej@norik.com>
 <4591cdd6-9a7b-cd1d-817d-8950c8976d10@roeck-us.net>
 <DB9PR10MB4652C8A69A6A3F38B93ED18880679@DB9PR10MB4652.EURPRD10.PROD.OUTLOOK.COM>
 <dcd75a82-5837-8d78-0a9f-6e5b7eafff28@roeck-us.net>
 <DB9PR10MB46521F419F43685F7103D32480679@DB9PR10MB4652.EURPRD10.PROD.OUTLOOK.COM>
 <ef3f98e1-6661-84ed-1bde-747b1330aba2@norik.com>
From:   Guenter Roeck <linux@roeck-us.net>
Message-ID: <8257f965-7f0f-f513-9d86-e43baf010967@roeck-us.net>
Date:   Tue, 30 Nov 2021 23:01:19 -0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.14.0
MIME-Version: 1.0
In-Reply-To: <ef3f98e1-6661-84ed-1bde-747b1330aba2@norik.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 11/30/21 10:42 PM, Andrej Picej wrote:
> 
> On 30. 11. 21 18:46, Adam Thomson wrote:
>> On 30 November 2021 16:40, Guenter Roeck wrote:
>>
>>>>> Why does it need a value ? Why not just bool ?
>>>>
>>>> One argument might be that if the property isn't provided then the OTP
>>>> configured value can persist without needing a FW change around this DT
>>> binding.
>>>>
>>>> My belief though is that the majority of users would have this property set to 0
>>>> by default in OTP, so a boolean would be OK I think here to enable watchdog
>>>> shutdown.
>>>>
>>>
>>> Sorry, you lost me.
>>>     dlg,wdt-sd = <0>;
>>> is the current situation, and identical to not having the property in
>>> the first place.
>>>     dlg,wdt-sd = <1>;
>>> is new. I don't see the difference to
>>>     dlg,wdt-sd;
>>> vs. not having the property at all (which is, again, the current situation).
>>> Since it has to be backward compatible,
>>>     dlg,wdt-sd = <0>;
>>> will always be identical to not having the property at all.
>>> I can not find a situation where an integer would have any benefits over a
>>> boolean.
>>
>> So if you have a binary DT binding, it's either there or it isn't which implies
>> the bit to be set to 0/1 in this case. If you have a binding which has a value,
>> there can be 3 outcomes in this discussion:
>>
>>   1) Binding = 0, bit is set to 0
>>   2) Binding = 1, bit is set to 1
>>   3) Binding NOT present in DT, OTP default value in HW remains untouched
>>
>> Say a platform updates to a later kernel version, but sticks with existing DT
>> FW (i.e. the new boolean binding isn't present in FW), then the following could
>> happen:
>>
>>   1) OTP for DA9061/2 has this bit set to 1, system expectation is that watchdog
>>      triggers SHUTDOWN.
>>   2) New driver checks existance of 'dlg,wdt-sd' but it's obviously not there so
>>      assumes the bit should be set to 0 and does so
>>   3) When the watchdog fires, it will no longer trigger SHUTDOWN but instead
>>      POWER-DOWN due to binary handling of new boolean binding.
>>
> 
> This was my thinking exactly. I also first thought about boolean value, but I then moved to the integer value of 0 or 1 after checking the OTP default for this bit. The da9062 I'm working with has the bit set to 1 by default.

That needs to be be documented.

Guenter