Re: [PATCH V1 1/1] watchdog: f71808e_wdt: fix F81866 bit operation

[Guenter Roeck <linux@roeck-us.net>]

On 3/24/19 11:10 PM, Ji-Ze Hong (Peter Hong) wrote:
> Guenter Roeck 於 2019/3/22 下午 09:06 寫道:
>> On 3/21/19 8:36 PM, Ji-Ze Hong (Peter Hong) wrote:
>>> Fix error bit operation in watchdog_start()
>>>
>>
>> Hmm ... does that mean it never worked ? Did you test it this time ?
> 
> Sorry for lacking test procedure. I had only test the functional (reset)
> , not to test the register value.
> 
> The F81866 PIN70 (WDTRST#/GPIO15) is default set to WDTRST# function and
> the old code only change register 27h bit(4) - PORT_4E_EN.
> 
> If the mainboard entry port is 4Eh, the old code is equal to nothing
> done, but when the mainboard entry port is 2Eh, this code will make the
> change from entry port 2Eh to 4Eh.
> 
> https://html.alldatasheet.com/html-pdf/459086/FINTEK/F81866AD/26531/119/F81866AD.html
> 
>> A secondary concern is that the watchdog doesn't _have_ to trigger WDTRST,
>> but may also trigger PWOK. But that may be a separate issue.
> 
> Out watchdog is only support WDTRST#.
> 
>> Please add:
>>
>> Fixes: 14b24a88a3660 ("watchdog: f71808e_wdt: Add F81866 support")
> 
> OK, I'll add to v2
> 
>>> diff --git a/drivers/watchdog/f71808e_wdt.c b/drivers/watchdog/f71808e_wdt.c
>>> index 9a1c761258ce..9129485732c7 100644
>>> --- a/drivers/watchdog/f71808e_wdt.c
>>> +++ b/drivers/watchdog/f71808e_wdt.c
>>> @@ -387,18 +387,17 @@ static int watchdog_start(void)
>>>       case f81866:
>>>           /* Set pin 70 to WDTRST# */
>>> -        superio_clear_bit(watchdog.sioaddr, SIO_F81866_REG_PORT_SEL,
>>> -                  BIT(3) | BIT(0));
>>> -        superio_set_bit(watchdog.sioaddr, SIO_F81866_REG_PORT_SEL,
>>> -                BIT(2));
>>> +        superio_clear_bit(watchdog.sioaddr, SIO_F81866_REG_PORT_SEL, 3);
>>> +        superio_clear_bit(watchdog.sioaddr, SIO_F81866_REG_PORT_SEL, 0);
>>> +        superio_set_bit(watchdog.sioaddr, SIO_F81866_REG_PORT_SEL, 2);
>>
>> Better use superio_inb()/superio_outb(). The above is (much) more expensive,
>> and  we have no real idea what the impact of changing one bit at a time may be.
> 
> Could I add a superio_mask_write(reg, mask, data) with v2 patch like
> following fintek driver ?
> https://elixir.bootlin.com/linux/latest/source/drivers/tty/serial/8250/8250_fintek.c#L113
> 

If you wish. but not necessary. For the fix it would be better to have only the fix.
You could introduce the function in a second patch and use it wherever read/modify/write
of a mask is used in the driver (I think it was used at least in one other place).

Guenter