Re: [PATCH 3/5] spi: spi-qpic: Add qpic spi nand driver support

[Md Sadre Alam]

On 2/15/2024 7:44 PM, Mark Brown wrote:
> On Thu, Feb 15, 2024 at 07:18:54PM +0530, Md Sadre Alam wrote:
>
> > +config SPI_QPIC_SNAND
> > +	tristate "QPIC SNAND controller"
> > +	default y
>
> Why is this driver so special it should be enabled by default?
  Sorry by mistake I kept this enabled by default, will change in next
  patch.
> > +	depends on ARCH_QCOM
>
> Please add an || COMPILE_TEST so this gets some build coverage.
 Ok
> > +	help
> > +	  QPIC_SNAND (QPIC SPI NAND) driver for Qualcomm QPIC controller.
> > +	  QPIC controller supports both parallel nand and serial nand.
> > +	  This config will enable serial nand driver for QPIC controller.
> > +
> >   config SPI_QUP
> >   	tristate "Qualcomm SPI controller with QUP interface"
> >   	depends on ARCH_QCOM || COMPILE_TEST
> > diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
> > index 4ff8d725ba5e..1ac3bac35007 100644
> > --- a/drivers/spi/Makefile
> > +++ b/drivers/spi/Makefile
> > @@ -153,6 +153,7 @@ obj-$(CONFIG_SPI_XTENSA_XTFPGA)		+= spi-xtensa-xtfpga.o
> >   obj-$(CONFIG_SPI_ZYNQ_QSPI)		+= spi-zynq-qspi.o
> >   obj-$(CONFIG_SPI_ZYNQMP_GQSPI)		+= spi-zynqmp-gqspi.o
> >   obj-$(CONFIG_SPI_AMD)			+= spi-amd.o
> > +obj-$(CONFIG_SPI_QPIC_SNAND)            += spi-qpic-snand.o
>
> Please keep this sorted.
Ok
> > --- /dev/null
> > +++ b/drivers/spi/spi-qpic-snand.c
> > @@ -0,0 +1,1025 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (c) 2023, Qualcomm Innovation Center, Inc. All rights reserved.
>
> Please make the entire comment a C++ one so things look more
> intentional.
Ok
> > +#define snandc_set_read_loc_first(snandc, reg, cw_offset, read_size, is_last_read_loc)	\
> > +snandc_set_reg(snandc, reg,			\
> > +	      ((cw_offset) << READ_LOCATION_OFFSET) |		\
> > +	      ((read_size) << READ_LOCATION_SIZE) |			\
> > +	      ((is_last_read_loc) << READ_LOCATION_LAST))
> > +
> > +#define snandc_set_read_loc_last(snandc, reg, cw_offset, read_size, is_last_read_loc)	\
> > +snandc_set_reg(snandc, reg,			\
> > +	      ((cw_offset) << READ_LOCATION_OFFSET) |		\
> > +	      ((read_size) << READ_LOCATION_SIZE) |			\
> > +	      ((is_last_read_loc) << READ_LOCATION_LAST))
>
> For type safety and legibility please write these as functions, mark
> them as static inline if needed.
Ok
> > +void snandc_set_reg(struct qcom_nand_controller *snandc, int offset, u32 val)
> > +{
> > +	struct nandc_regs *regs = snandc->regs;
> > +	__le32 *reg;
> > +
> > +	reg = offset_to_nandc_reg(regs, offset);
> > +
> > +	if (reg)
> > +		*reg = cpu_to_le32(val);
> > +}
>
> This silently ignores writes to invalid registers, that doesn't seem
> great.
Ok
> > +	return snandc->ecc_stats.failed ? -EBADMSG : snandc->ecc_stats.bitflips;
>
> For legibility please just write normal conditional statements.
Ok
> > +static int qpic_snand_program_execute(struct qcom_nand_controller *snandc,
> > +				      const struct spi_mem_op *op)
> > +{
>
> > +       int num_cw = 4;
>
> > +	data_buf = (u8 *)snandc->wbuf;
>
> Why the cast?  If it's needed that smells like it's masking a bug, it
> looks like it's casting from a u8 * to a u8 *.
Ok Will fix in next patch.
> > +	for (i = 0; i < num_cw; i++) {
> > +		int data_size;
>
> All these functions appear to hard code "num_cw" to 4.  What is "num_cw"
> and why are we doing this per function?
QPIC controller internally works on code word size not on page and each
code word size is 512-bytes so if page size is 2K then num_cw = 4, if page
size is 4K then num_cw = 8.
Will not hard code this value to 4 or 8 , will fix this in next patch.
> > +static int qpic_snand_program_execute(struct qcom_nand_controller *snandc,
> > +                                     const struct spi_mem_op *op)
>
> > +	if (op->cmd.opcode == SPINAND_READID) {
> > +		snandc->buf_count = 4;
> > +		read_reg_dma(snandc, NAND_READ_ID, 1, NAND_BAM_NEXT_SGL);
> > +
> > +		ret = submit_descs(snandc);
> > +		if (ret)
> > +			dev_err(snandc->dev, "failure in submitting descriptor for readid\n");
> > +
> > +		nandc_read_buffer_sync(snandc, true);
> > +		memcpy(op->data.buf.in, snandc->reg_read_buf, snandc->buf_count);
>
> These memcpy()s don't seem great, why aren't we just reading directly
> into the output buffer?
  This reg_read_buf is being used in common API so that it will be used by both
  serial nand as well raw nand, so I can't directly use the output buffer since
  internally CW mechanism I have to maintain in common API.
> > +	if (op->cmd.opcode == SPINAND_GET_FEATURE) {
>
> This function looks like it should be a switch statement.
Ok
> > +static bool qpic_snand_is_page_op(const struct spi_mem_op *op)
> > +{
>
> > +	if (op->data.dir == SPI_MEM_DATA_IN) {
> > +		if (op->addr.buswidth == 4 && op->data.buswidth == 4)
> > +			return true;
> > +
> > +		if (op->addr.nbytes == 2 && op->addr.buswidth == 1)
> > +			return true;
> > +
> > +	} else if (op->data.dir == SPI_MEM_DATA_OUT) {
> > +		if (op->data.buswidth == 4)
> > +			return true;
> > +		if (op->addr.nbytes == 2 && op->addr.buswidth == 1)
> > +			return true;
> > +	}
>
> Again looks like a switch statement.
Ok
> > +	ctlr = devm_spi_alloc_master(dev, sizeof(*snandc));
> > +	if (!ctlr)
> > +		return -ENOMEM;
>
> Please use _alloc_controller.
Ok
> > +static int qcom_snand_remove(struct platform_device *pdev)
> > +{
> > +	struct spi_controller *ctlr = platform_get_drvdata(pdev);
> > +
> > +	spi_unregister_controller(ctlr);
> > +
> > +	return 0;
> > +}
>
> We don't disable any of the clocks in the remove path.
OK will fix in next patch.

Thanks for reviewing, will address all your comments in the next patch.

Regards,
Alam.