Re: [PATCH v2 3/4] watchdog: add meson secure watchdog driver

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

On 10/21/19 1:03 AM, Xingyu Chen wrote:
> Hi, Guenter
> 
> On 2019/10/21 0:56, Guenter Roeck wrote:
>> On 10/18/19 1:33 AM, Xingyu Chen wrote:
>>> The watchdog controller on the Meson-A/C series SoCs is moved to secure
>>> world, watchdog operation needs to be done in secure EL3 mode via ATF,
>>> Non-secure world can call SMC instruction to trap to AFT for watchdog
>>> operation.
>>>
>>> Signed-off-by: Xingyu Chen <xingyu.chen@amlogic.com>
>>> ---
>>>   drivers/watchdog/Kconfig         |  17 ++++
>>>   drivers/watchdog/Makefile        |   1 +
>>>   drivers/watchdog/meson_sec_wdt.c | 187 +++++++++++++++++++++++++++++++++++++++
>>>   3 files changed, 205 insertions(+)
>>>   create mode 100644 drivers/watchdog/meson_sec_wdt.c
>>>
>>> diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
>>> index 58e7c10..e84be42 100644
>>> --- a/drivers/watchdog/Kconfig
>>> +++ b/drivers/watchdog/Kconfig
>>> @@ -826,6 +826,23 @@ config MESON_GXBB_WATCHDOG
>>>         To compile this driver as a module, choose M here: the
>>>         module will be called meson_gxbb_wdt.
>>> +config MESON_SEC_WATCHDOG
>>> +    tristate "Amlogic Meson Secure watchdog support"
>>> +    depends on MESON_SM
>>> +    depends on ARCH_MESON || COMPILE_TEST
>>
>> This dependency is pointless. MESON_SM already depends on ARCH_MESON,
>> thus specifying "COMPILE_TEST" here adds no value but only
>> creates confusion.
> Thanks for your analysis, perhaps i should remove the line below.
> - depends on ARCH_MESON || COMPILE_TEST
> 
> Is it ok to modify code above like this ?

Yes.
[ ... ]

>>> +static unsigned int meson_sec_wdt_get_timeleft(struct watchdog_device *wdt_dev)
>>> +{
>>> +    int ret;
>>> +    unsigned int timeleft;
>>> +    struct meson_sec_wdt *data = watchdog_get_drvdata(wdt_dev);
>>> +
>>> +    ret = meson_sm_call(data->fw, SM_WATCHDOG_OPS, &timeleft,
>>> +                MESON_SIP_WDT_GETTIMELEFT, 0, 0, 0, 0);
>>> +
>>> +    if (ret)
>>> +        return ret;
>>
>> Meh, that doesn't work. I just realized that the return type is unsigned,
>> so returning a negative error code is pointless. Guess we'll have to
>> live with returning 0 in this case after all. I wonder if we should
>> fix the API and return an integer (with negative error code), but that
>> is a different question.
> Thanks for your review.
> 
> IMO, if returning an integer, and the value which copy to user buf should be formatted with %d instead of %u (see timeleft_show), it will cause the max value of timeleft is reduced from 4294967295 to 2147483647. but i'am not sure whether it will bring risk.

Not that it matters right now, but I don't think that limiting 'timeleft'
reporting to 2147483647 seconds, or ~68 years, would cause any risk.
It would just be a large patch changing several drivers all at once,
that is all.

> 
> So i also think returning 0 may be better in this case.

Yes, please do that.

Thanks,
Guenter