Re: [PATCH 01/12] thinkpad_ec: New driver for ThinkPad embedded controller access

["Shem Multinymous" <multinymous@gmail.com>]

Hi Pavel,

Thanks for the sign-offs!

On 8/7/06, Pavel Machek <pavel@suse.cz> wrote:
> > +module_param_named(debug, tp_debug, int, 0600);
> > +MODULE_PARM_DESC(debug, "Debug level (0=off, 1=on)");
> > +
> > +/* A few macros for printk()ing: */
> > +#define DPRINTK(fmt, args...) \
> > +  do { if (tp_debug) printk(KERN_DEBUG fmt, ## args); } while (0)
>
> Is not there generic function doing this?

None that I found. Many drivers do it this way.


> > +EXPORT_SYMBOL_GPL(thinkpad_ec_lock);
> > +EXPORT_SYMBOL_GPL(thinkpad_ec_try_lock);
> > +void thinkpad_ec_unlock(void)
> > +{
> > +     up(&thinkpad_ec_mutex);
> > +}
> > +
>
> Do we need these wrappers? Perhaps just directly exporting the mutex?

Yes, we may end up needing to lock away other systems (ACPI?) that
touch the same ports. Apparently not an issue right now, but could
change with new firmware.
Also, that's what Alan Cox told me to do. :-)



> > +     for (i=0; i<TPC_REQUEST_RETRIES; ++i) {
>
> I'd write i++ here (and in other loops)... just for consistency with
> rest of kernel.

OK.
"++i" is the recommended form for C++, got used to that.


> > +     struct dmi_device *dev = NULL;
>
> unneeded initializer.

On a local variable?!


> > +static int __init thinkpad_ec_init(void)
> > +{
> > +     if (!check_dmi_for_ec()) {
> > +             printk(KERN_ERR "thinkpad_ec: no ThinkPad embedded controller!\n");
> > +             return -ENODEV;
>
> KERN_ERR is little strong here, no?

Not sure what's the right one. The user tried to load a module and the
module can't do that; I saw some drivers use KERN_ERR some
KERN_WARNING in similar cases. Is there some guideline on choosing
printk levels?

OK on all other points.

  Shem