On Sun, Apr 12, 2020 at 02:28:31PM -0500, Sanjay R Mehta wrote:

> +++ b/drivers/spi/spi-amd.c
> @@ -0,0 +1,341 @@
> +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
> +/*
> + * AMD SPI controller driver
> + *

Please make the entire comment a C++ one so things look more
intentional.

> +#define DRIVER_NAME		"amd_spi"

This is unused.

> +/* M_CMD OP codes for SPI */
> +#define SPI_XFER_TX		1
> +#define SPI_XFER_RX		2

These constants should be namespaced, they're likely to collide with
generic additions.

> +static void amd_spi_execute_opcode(struct spi_master *master)
> +{
> +	struct amd_spi *amd_spi = spi_master_get_devdata(master);
> +	bool spi_busy;
> +
> +	/* Set ExecuteOpCode bit in the CTRL0 register */
> +	amd_spi_setclear_reg32(master, AMD_SPI_CTRL0_REG, AMD_SPI_EXEC_CMD,
> +			       AMD_SPI_EXEC_CMD);
> +
> +	/* poll for SPI bus to become idle */
> +	spi_busy = (ioread32((u8 __iomem *)amd_spi->io_remap_addr +
> +		    AMD_SPI_CTRL0_REG) & AMD_SPI_BUSY) == AMD_SPI_BUSY;
> +	while (spi_busy) {
> +		set_current_state(TASK_INTERRUPTIBLE);
> +		schedule();
> +		set_current_state(TASK_RUNNING);
> +		spi_busy = (ioread32((u8 __iomem *)amd_spi->io_remap_addr +
> +			    AMD_SPI_CTRL0_REG) & AMD_SPI_BUSY) == AMD_SPI_BUSY;
> +	}

This is a weird way to busy wait - usually you'd use a cpu_relax()
rather than a schedule().  There's also no timeout here so we could busy
wait for ever if something goes wrong.

> +static int amd_spi_master_setup(struct spi_device *spi)
> +{
> +	struct spi_master *master = spi->master;
> +	struct amd_spi *amd_spi = spi_master_get_devdata(master);
> +
> +	amd_spi->chip_select = spi->chip_select;
> +	amd_spi_select_chip(master);

This looks like it will potentially affect devices other than the
current one.  setup() may be called while other devices are active it
shouldn't do that.

> +		} else if (m_cmd & SPI_XFER_RX) {
> +			/* Store no. of bytes to be received from
> +			 * FIFO
> +			 */
> +			rx_len = xfer->len;
> +			buffer = (u8 *)xfer->rx_buf;

> +		/* Read data from FIFO to receive buffer  */
> +		for (i = 0; i < rx_len; i++)
> +			buffer[i] = ioread8((u8 __iomem *)amd_spi->io_remap_addr
> +					    + AMD_SPI_FIFO_BASE
> +					    + tx_len + i);

This will only work for messages with a single receive transfer, if
there are multiple transfers then you'll need to store multiple buffers
and their lengths.

> +static int amd_spi_master_transfer(struct spi_master *master,
> +				   struct spi_message *msg)
> +{
> +	struct amd_spi *amd_spi = spi_master_get_devdata(master);
> +
> +	/*
> +	 * Extract spi_transfers from the spi message and
> +	 * program the controller.
> +	 */
> +	amd_spi_fifo_xfer(amd_spi, msg);
> +
> +	return 0;
> +}

This function is completely redundant, just inline amd_spi_fifo_xfer().
It also ignores all errors which isn't great.

> +	/* Initialize the spi_master fields */
> +	master->bus_num = 0;
> +	master->num_chipselect = 4;
> +	master->mode_bits = 0;
> +	master->flags = 0;

This device is single duplex so should flag that.

> +	err = spi_register_master(master);
> +	if (err) {
> +		dev_err(dev, "error registering SPI controller\n");
> +		goto err_iounmap;

It's best to print the error code to help people debug things.