On 02/27/2018 05:00 AM, Wolfram Sang wrote:
- rwdt_write(priv, 0, RWTCSRB);
- rwdt_write(priv, priv->cks, RWTCSRA);
Isn't this done implicitly with the above already ?
After all, priv->cks won't have RWTCSRA_TME set.
Yes. The request came from a group doing some (safety?) audit who didn't
like the implicit handling which might change if the code in probe()
might change somewhen. And the datasheet explicitly says to disable the
timer first before doing anything with the clock dividers. Given that, I
can agree to this change, too.
Defensive code is fine, bit that is a bit too defensive. We might as well
start checking value ranges in the drivers' set_timeout functions just in case
the core gets it wrong. Just add a note to the probe function where cks is
initialized that RWTCSRA_TME must not be set, and that RWTCSRA_TME must be
cleared before changing the divider.
Guenter
