Received-SPF: pass (google.com: domain of linux-kernel+bounces-44398-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Date: Tue, 30 Jan 2024 12:30:23 +0200
From: Raag Jadav <raag.jadav@intel.com>
To: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: u.kleine-koenig@pengutronix.de, jarkko.nikula@linux.intel.com,
	mika.westerberg@linux.intel.com, lakshmi.sowjanya.d@intel.com,
	linux-pwm@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel
 Elkhart Lake
Message-ID: <ZbjPv_-S-6CQsaja@black.fi.intel.com>
References: <20240122030238.29437-1-raag.jadav@intel.com>
 <20240122030238.29437-2-raag.jadav@intel.com>
 <ZbZqZDvdw-_D3hyb@smile.fi.intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZbZqZDvdw-_D3hyb@smile.fi.intel.com>

On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> > Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> > function with 8 channels each. Add support for the remaining channels.
> 
> ...
> 
> > +static int dwc_pwm_init(struct device *dev, const struct dwc_pwm_info *info, void __iomem *base)
> > +{
> > +	/* Default values for single instance devices */
> > +	unsigned int nr = 1, size = 0;
> > +	int i, ret;
> > +
> > +	/* Fill up values from driver_data, if any */
> > +	if (info) {
> > +		nr = info->nr;
> > +		size = info->size;
> > +	}
> > +
> > +	for (i = 0; i < nr; i++) {
> > +		struct dwc_pwm *dwc;
> > +
> > +		dwc = dwc_pwm_alloc(dev);
> > +		if (!dwc)
> > +			return -ENOMEM;
> > +
> > +		dwc->base = base + (i * size);
> > +
> > +		ret = devm_pwmchip_add(dev, &dwc->chip);
> > +		if (ret)
> > +			return ret;
> > +	}
> > +
> > +	return 0;
> > +}
> 
> Why not doing this slightly differently?
> 
> First option: Always provide driver data (info is never NULL).

Allowing empty driver_data would save us from adding dummy info
for single instance devices in the future.

> Second option, have the body of the for-loop be factored to a helper
> dwc_pwm_init_one() and here
> 
> 	if (!info)
> 		return dwc_pwm_init_one(..., 1, 0);
> 
> 	for (i = 0; i < info->nr; i++) {
> 		ret = dwc_pwm_init_one(...);
> 		if (ret)
> 			return ret;
> 	}

Considering above, we're looking at something like this.

static int dwc_pwm_init_one(struct device *dev, void __iomem *base, unsigned int size)
{
        struct dwc_pwm *dwc;

        dwc = dwc_pwm_alloc(dev);
        if (!dwc)
                return -ENOMEM;

        dwc->base = base + size;

        return devm_pwmchip_add(dev, &dwc->chip);
}

	...

        if (info) {
                for (i = 0; i < info->nr; i++) {
                        ret = dwc_pwm_init_one(dev, base, i * info->size);
                        if (ret)
                                return ret;
                }
        } else {
                ret = dwc_pwm_init_one(dev, base, 0);
                if (ret)
                        return ret;
        }
	...

Raag