Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 1 May 2023 16:04:31 +0100
From:   Jonathan Cameron <jic23@kernel.org>
To:     Matti Vaittinen <mazziesaccount@gmail.com>
Cc:     Mehdi Djait <mehdi.djait.k@gmail.com>,
        krzysztof.kozlowski+dt@linaro.org,
        andriy.shevchenko@linux.intel.com, robh+dt@kernel.org,
        lars@metafoo.de, linux-iio@vger.kernel.org,
        linux-kernel@vger.kernel.org, devicetree@vger.kernel.org
Subject: Re: [PATCH v3 7/7] iio: accel: Add support for Kionix/ROHM
 KX132-1211 accelerometer
Message-ID: <20230501160431.602e3086@jic23-huawei>
In-Reply-To: <00905352-f9a4-a8bf-63b7-92e3b1d2a4f4@gmail.com>
References: <cover.1682373451.git.mehdi.djait.k@gmail.com>
        <593798a44c8ba45f969b86aa29e172d59065958c.1682373451.git.mehdi.djait.k@gmail.com>
        <00905352-f9a4-a8bf-63b7-92e3b1d2a4f4@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk


> > +static int kx132_get_fifo_bytes(struct kx022a_data *data)
> > +{
> > +	struct device *dev = regmap_get_device(data->regmap);
> > +	__le16 buf_status;
> > +	int ret, fifo_bytes;
> > +
> > +	ret = regmap_bulk_read(data->regmap, data->chip_info->buf_status1,
> > +			       &buf_status, sizeof(buf_status));
> > +	if (ret) {
> > +		dev_err(dev, "Error reading buffer status\n");
> > +		return ret;
> > +	}
> > +
> > +	fifo_bytes = le16_to_cpu(buf_status);
> > +	fifo_bytes &= data->chip_info->buf_smp_lvl_mask;  
> 
> This is probably just my limitation but I've hard time thinking how this 
> works out on BE machines. It'd be much easier for me to understand this 
> if the data was handled as two u8 values and mask was applied before 
> endianes conversion. (Eg - untested pseudo code follows;
> 
> __le16 buf_status;
> u8 *reg_data;
> 
> ...
> 
> ret = regmap_bulk_read(data->regmap, data->chip_info->buf_status1,
> 			&buf_status, sizeof(buf_status));
> ...
> 
> reg_data = (u8 *)&buf_status;
> 
> /* Clear the unused bits form 2.nd reg */
> reg_data[1] = reg_data[i] & MASK_SMP_LVL_REG_HIGH_BITS;
> 
> /* Convert to CPU endianess */
> fifo_bytes = le16_to_cpu(buf_status);
> 
> Well, others may have different view on this :)

:) 

I go the other way. It's less obvious to me that it is appropriate
to apply le16_to_cpu(buf_status) after applying a mask to some
bits. The moment that is appropriate, then we certainly hope a single
mask application is as well.

I think treating it as a 16 bit register is appropriate, in particular
as the field is described as SMP_LEV[9:0] on the datasheet
(of course there are datasheets that do that for unconnected sets of
bits so this doesn't always work ;)

Jonathan