On Mon, Aug 20, 2012 at 05:37:48AM +0000, Arnd Bergmann wrote:
> On Monday 20 August 2012, Kim, Milo wrote:
> > +#ifdef CONFIG_PWM
> > +static int lm3530_pwm_request(struct lm3530_data *drvdata)
> > +{
> > + int pwm_id;
> > +
> > + /* if the pwm device exists, skip requesting the device */
> > + if (drvdata->pwm)
> > + return 0;
> > +
> > + pwm_id = drvdata->pdata ? drvdata->pdata->pwm_id : 0;
> > +
> > + drvdata->pwm = pwm_request(pwm_id, "lm3530-pwm");
> > + drvdata->period_ns = drvdata->pdata ? drvdata->pdata->period_ns : 0;
> > +
> > + return IS_ERR(drvdata->pwm) ? PTR_ERR(drvdata->pwm) : 0;
> > +}
> > +
>
> A few comments on this:
>
> * Rather than having to do the #ifdef here, I think it would be better if
> the PWM subsystem provided stub functions for pwm_request, pwm_config,
> pwm_enable, pwm_disable and pwm_free that do nothing, so you can in effect
> let the compiler optimize away the above code.That's actually on my TODO list, but I think it needs to wait until we have gotten rid of all legacy implementations. The stubs would have to move into the !CONFIG_PWM branch, which will in turn break because the legacy implementations would provide non-inlined duplicates. Thierry
pgpVvLt2WGcKj.pgp
Description: PGP signature
