Re: [PATCH 3/9] watchdog: Add Photonicat PMU watchdog driver

[Junhao Xie <bigfoot@classfun.cn>]

On 2024/9/6 19:52, Guenter Roeck wrote:
> On 9/6/24 02:36, Junhao Xie wrote:
>> This driver provides access to Photonicat PMU watchdog functionality.
>>
[...]
>> +
>> +struct pcat_watchdog {
>> +    struct device *dev;
> 
> I don't see what this is used for.

I used to use this for logging, but now they are gone, I will delete it.

> 
[...]
>> +
>> +static int pcat_wdt_setup(struct pcat_watchdog *data, int timeout)
>> +{
>> +    int ret;
>> +    u8 time = 0;
> 
> Unnecessary initialization.
> 
>> +    u8 times[3] = { 60, 60, 0 };
>> +
>> +    time = MIN(255, MAX(0, timeout));
>> +
>> +    ret = pcat_pmu_write_data(data->pmu, PCAT_CMD_WATCHDOG_TIMEOUT_SET,
>> +                  times, sizeof(times));
> 
> Where does this actually send the timeout to the chip ?
> 

I forgot to fill in timeout into times[2] during refactoring process, I will fix it.

>> +    if (!ret)
[...]>> +
>> +static int pcat_wdt_set_timeout(struct watchdog_device *wdev, unsigned int val)
>> +{
>> +    int ret = 0;
>> +    struct pcat_watchdog *data = watchdog_get_drvdata(wdev);
>> +
>> +    data->timeout = val;
> 
> This needs to store 'timeout' in wdev. Storing it locally is unnecessary.
> 
>> +    if (data->started)
>> +        ret = pcat_wdt_setup(data, data->timeout);
> 
> This is misleading because it would permit setting the timeout to
> 0 when the watchdog isn't running, and then when the watchdog is started
> it would not really start it. The code should not use a local "started"
> variable but call watchdog_active(). It should also not accept "0"
> as a valid timeout.
> 

I will fix the pcat_wdt_set_timeout.

>> +
[...]
>> +
>> +    watchdog->dev = dev;
>> +    watchdog->pmu = dev_get_drvdata(dev->parent);
>> +    watchdog->wdd.info = &pcat_wdt_info;
>> +    watchdog->wdd.ops = &pcat_wdt_ops;
>> +    watchdog->wdd.timeout = 60;
>> +    watchdog->wdd.max_timeout = U8_MAX;
>> +    watchdog->wdd.min_timeout = 0;
> 
> This effectively lets the user ... kind of ... stop the watchdog
> by setting the timeout to 0. This is not acceptable.
> 
>> +    watchdog->wdd.parent = dev;
>> +
>> +    watchdog_stop_on_reboot(&watchdog->wdd);
>> +    watchdog_set_drvdata(&watchdog->wdd, watchdog);
>> +    platform_set_drvdata(pdev, watchdog);
>> +
> No watchdog_init_timeout() ?

Thanks for your correction, I will fix it.

> 
>> +    return devm_watchdog_register_device(dev, &watchdog->wdd);
[...]
>> +MODULE_LICENSE("GPL");
> 

Thanks for your review, I will fix all problems in next version!

Best regards,
Junhao