[PATCH v2 3/3] mxs: added driver for OCOTP in i.MX23 and i.MX28

[Christoph G. Baumann]

Hello Sascha,

thanks for your input. I must admit that I didn't do a thorough review of the
FSL code...


> Sascha Hauer <s.hauer at pengutronix.de> hat am 17. Juli 2013 um 21:26
> geschrieben:
> [...]
> > +static DEFINE_MUTEX(otp_mutex);
> > +static struct kobject *otp_kobj;
> > +static struct attribute **attrs;
> > +static struct kobj_attribute *kattr;
> > +static struct attribute_group attr_group;
> > +static unsigned long otp_hclk_saved;
> > +static u32 otp_voltage_saved;
> > +static resource_size_t otp_base_address;
> > +struct regulator *regu;
>
> Collect your driver data in a struct and pass it around in functions
> calls. No need to limit this driver to a single instance.

The OTP device is unique in the CPU and several instances i.e. potential
concurrent access is considered harmful. So I kept this construction
from FSL.


> > +static int otp_write_prepare(void)
> > +{
> > +     struct clk *hclk;
> > +     int err = 0;
> > +
> > +     /* [1] HCLK to 24MHz. */
> > +     hclk = clk_get_sys("hbus", NULL);
>
> This is a driver, so you should use regular clk_get, not clk_get_sys.
> Besides, this has to be done during probe. clk_get takes a reference to
> the clock which has to be released.

As far as I remember testing clk_get didn't work in this place.
In the API doc I only found the warning:
"clk_get_sys should not be called from within interrupt context.".


Regards,
Christoph