Re: [PATCH 2/2] SPI: Add support for Zynq Quad SPI controller

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Apr 03, 2014 at 10:33:07PM +0530, Punnaiah Choudary Kalluri wrote:

Overall this looks fairly good, there are a few issues that need to be
looked at but they're not too invasive.  Please also check for coding
style issues, quite a few spaces before commas for example.

> +/*
> + * The modebits configurable by the driver to make the SPI support different
> + * data formats
> + */
> +#define MODEBITS			(SPI_CPOL | SPI_CPHA)

This should be namespaced and seems generally useful - please add it to
the header if there's no suitable definition already.

> +/**
> + * zynq_qspi_copy_read_data - Copy data to RX buffer
> + * @xqspi:	Pointer to the zynq_qspi structure
> + * @data:	The 32 bit variable where data is stored
> + * @size:	Number of bytes to be copied from data to RX buffer
> + */
> +static void zynq_qspi_copy_read_data(struct zynq_qspi *xqspi, u32 data, u8 size)
> +{
> +	if (xqspi->rxbuf) {
> +		memcpy(xqspi->rxbuf, ((u8 *) &data) + 4 - size, size);
> +		xqspi->rxbuf += size;
> +	}
> +	xqspi->bytes_to_receive -= size;
> +}

Does this and the write function really need to be a separate function -
it's trivial and used once?  It's probably more beneficial to split out
some of the more complex logic later on that's causing the indentation
to get too deep.

> +/**
> + * zynq_prepare_transfer_hardware - Prepares hardware for transfer.
> + * @master:	Pointer to the spi_master structure which provides
> + *		information about the controller.
> + *
> + * This function enables SPI master controller.
> + *
> + * Return:	Always 0
> + */
> +static int zynq_prepare_transfer_hardware(struct spi_master *master)
> +{
> +	struct zynq_qspi *xqspi = spi_master_get_devdata(master);
> +
> +	clk_enable(xqspi->devclk);
> +	clk_enable(xqspi->aperclk);

Not clk_prepare_enable()?  You need to check for errors too.

> +static int zynq_qspi_setup_transfer(struct spi_device *qspi,
> +				    struct spi_transfer *transfer)
> +{
> +	struct zynq_qspi *xqspi = spi_master_get_devdata(qspi->master);
> +	u32 config_reg, req_hz, baud_rate_val = 0;
> +
> +	if (transfer)
> +		req_hz = transfer->speed_hz;
> +	else
> +		req_hz = qspi->max_speed_hz;

Why would a transfer be being set up without a transfer being provided?

> +/**
> + * zynq_qspi_setup - Configure the QSPI controller
> + * @qspi:	Pointer to the spi_device structure
> + *
> + * Sets the operational mode of QSPI controller for the next QSPI transfer, baud
> + * rate and divisor value to setup the requested qspi clock.
> + *
> + * Return:	0 on success and error value on failure
> + */
> +static int zynq_qspi_setup(struct spi_device *qspi)
> +{
> +	if (qspi->master->busy)
> +		return -EBUSY;
> +
> +	return zynq_qspi_setup_transfer(qspi, NULL);
> +}

No, this is broken - you have to support setup() while the hardware is
active.  Just remove this if there's nothing to do and set up on the
transfer.

> +	intr_status = zynq_qspi_read(xqspi, ZYNQ_QSPI_STATUS_OFFSET);
> +	zynq_qspi_write(xqspi, ZYNQ_QSPI_STATUS_OFFSET , intr_status);

Coding style.

> +				if (xqspi->rxbuf) {
> +					(*(u32 *)xqspi->rxbuf) =
> +					zynq_qspi_read(xqspi,
> +						       ZYNQ_QSPI_RXD_OFFSET);
> +					xqspi->rxbuf += 4;

This only works in 4 byte words?  That seems a bit limited.
Alternatively, if it works with smaller sizes (as it appears to) then
isn't this at risk of overflowing buffers?

> +static int __maybe_unused zynq_qspi_suspend(struct device *_dev)
> +{
> +	struct platform_device *pdev = container_of(_dev,
> +			struct platform_device, dev);
> +	struct spi_master *master = platform_get_drvdata(pdev);
> +
> +	spi_master_suspend(master);
> +
> +	zynq_unprepare_transfer_hardware(master);

Why are you unpreparing the hardware - the framework should be doing
that for you if the device is active, if it's not you've got an extra
clock disable here?

> +static int __maybe_unused zynq_qspi_resume(struct device *dev)

This doesn't appear to be calling init_hw() - is it guaranteed that all
the register settings written there are OK after power on?

> +	ret = clk_prepare_enable(xqspi->aperclk);
> +	if (ret) {
> +		dev_err(&pdev->dev, "Unable to enable APER clock.\n");
> +		goto remove_master;
> +	}
> +
> +	ret = clk_prepare_enable(xqspi->devclk);
> +	if (ret) {
> +		dev_err(&pdev->dev, "Unable to enable device clock.\n");
> +		goto clk_dis_aper;
> +	}

The driver isn't using runtime_pm or otherwise disabling the clocks when
idle so it looks like the clocks will always be enabled and the
management in prepare and unprepare won't have any practical effect.  I
can see needing at least one of the clocks for setting up the device but
probably either the probe should disable them as it finishes or you
should move the clock enable/disable from prepare/unprepare to runtime
PM.

Attachment: signature.asc
Description: Digital signature


[Index of Archives]     [Linux Kernel]     [Linux ARM (vger)]     [Linux ARM MSM]     [Linux Omap]     [Linux Arm]     [Linux Tegra]     [Fedora ARM]     [Linux for Samsung SOC]     [eCos]     [Linux Fastboot]     [Gcc Help]     [Git]     [DCCP]     [IETF Announce]     [Security]     [Linux MIPS]     [Yosemite Campsites]

  Powered by Linux