Re: [PATCH v3 07/10] rtc: max77686: Use dev_warn() instead of pr_warn()

[Alexandre Belloni]

Hi,

On 27/01/2016 at 11:05:36 +0900, Krzysztof Kozlowski wrote :
> On 27.01.2016 10:53, Javier Martinez Canillas wrote:
> > Hello Andi,
> > 
> > Thanks a lot for your feedback and review.
> > 
> > On 01/26/2016 10:22 PM, Andi Shyti wrote:
> >> Hi Javier,
> >>
> >>>           if (tm->tm_year < 100) {
> >>> -            pr_warn("RTC can't handle year %d. Assume it's 2000.\n",
> >>> -                1900 + tm->tm_year);
> >>> +            dev_warn(info->dev,
> >>> +                 "RTC can't handle year %d. Assume it's 2000\n",
> >>> +                 1900 + tm->tm_year);
> >>>               return -EINVAL;
> >>
> >> Because we are returning an error value, why not use dev_err()?
> >>
> > 
> > You are absolutely right. Since the driver was using pr_warn(), I used
> > dev_warn() but dev_err() would had been correct.
> 
> Wait. The message says that "2000 will be assumed" which is not an
> error. The message indicates that driver will proceed, thus the warning.
> 
> However the driver won't proceed because the max77686_rtc_set_time()
> will abort. This came from max8997 which has the same issue.
> 
> This means that either message should be changed (dev_err() without the
> "assume" verb) or the function should not abort and set the year to
> 2000+something (then dev_warn()... look at rtc-ds3234.c and rtc-mcp795.c).
> 
> The easiest would be to choose #1 - no changes in the logic.
> 

My stance on that is to never set a date that differs from the requested
date. Else, userspace has no way of knowing whether this is an erroneous
date or the real date when reading back.

I think I had a look and the driver is already doing the right thing but
the message is wrong.

-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com