RE: [PATCH V3 3/3] mfd: da9063: MFD support for OnKey driver

[Opensource [Steve Twiss]]

On 04 May 2015 08:47, Geert Uytterhoeven wrote:

> On Thu, Apr 30, 2015 at 1:26 PM, S Twiss <stwiss.opensource@xxxxxxxxxxx>
> wrote:
> > Add MFD support for the DA9063 OnKey driver
> 
> Thanks  for your patch!
> 
> > +static int da9063_clear_fault_log(struct da9063 *da9063)
> > +{	
> > +       int ret = 0;
> > +       int fault_log = 0;
> > +
> > +       ret = regmap_read(da9063->regmap, DA9063_REG_FAULT_LOG,
> &fault_log);
> > +       if (ret < 0) {
> > +               dev_err(da9063->dev, "Cannot read FAULT_LOG.\n");
> > +               return -EIO;
> > +       }
> > +
> > +       if (fault_log) {
> > +               if (fault_log & DA9063_TWD_ERROR)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_TWD_ERROR\n");
> > +               if (fault_log & DA9063_POR)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_POR\n");
> > +               if (fault_log & DA9063_VDD_FAULT)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_VDD_FAULT\n");
> > +               if (fault_log & DA9063_VDD_START)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_VDD_START\n");
> > +               if (fault_log & DA9063_TEMP_CRIT)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_TEMP_CRIT\n");
> > +               if (fault_log & DA9063_KEY_RESET)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_KEY_RESET\n");
> > +               if (fault_log & DA9063_NSHUTDOWN)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_NSHUTDOWN\n");
> > +               if (fault_log & DA9063_WAIT_SHUT)
> > +                       dev_dbg(da9063->dev,
> > +                               "Fault log entry detected: DA9063_WAIT_SHUT\n");
> > +       }
> > +
> > +       ret = regmap_write(da9063->regmap,
> > +                          DA9063_REG_FAULT_LOG,
> > +                          fault_log);
> > +       if (ret < 0)
> > +               dev_err(da9063->dev,
> > +                       "Cannot reset FAULT_LOG values %d\n", ret);
> > +
> > +       return ret;
> > +}
> > +
> >  int da9063_device_init(struct da9063 *da9063, unsigned int irq)
> >  {
> >         struct da9063_pdata *pdata = da9063->dev->platform_data;
> >         int model, variant_id, variant_code;
> >         int ret;
> >
> > +       ret = da9063_clear_fault_log(da9063);
> > +       if (ret < 0)
> > +               dev_err(da9063->dev, "Cannot clear fault log\n");
> > +
> >         if (pdata) {
> >                 da9063->flags = pdata->flags;
> >                 da9063->irq_base = pdata->irq_base;
> 
> The code above doesn't seem to match the patch description.
> Can you please explain why it is needed?
> 

Hi Geert,

Yes, at first it does seem that the fault-log clearing function is unrelated to the
"MFD support for the DA9063 OnKey driver", but there is an important connection
that makes it essential for the OnKey driver in this case.

I have made some explanation of the OnKey's operation in this thread here:
https://lkml.org/lkml/2015/4/29/406

But I only described the OnKey's point-of-view for case (4) of the "OnKey operation"
Case (4): the long-long press and no key release -- the hardware power-cut when
software is unable to respond (for whatever reason).

In this case, if the software was not responsive and a hardware shutdown of the
device was chosen, the FAULT_LOG would persist with information that would be
accessible when the device was restarted: it would be possible to take action the
next time the device was turned on again (maybe to trigger some mitigation exercise
in the bootloader -- e.g. say to complete memory checks or put the device into a
safe mode). 

However this mitigation exercise and clearance of the fault-log cannot be counted on 
outside the Linux kernel and so the clearance function da9063_clear_fault_log () has
been done here.

If this clearance function did not exist, then after a hardware shutdown (due to S/W
failure), the next time the device is restarted the FAULT_LOG would persist with values
from the previous error.

Regards,
Steve