Re: [PATCH 5/5] mtd: spi-nor: Add Synopsys DesignWare Octal SPI driver

[Jules Maselbas <jmaselbas@zdiv.net>]

Hi Sascha,

On Wed, Jul 26, 2023 at 01:39:12PM +0200, Sascha Hauer wrote:
> Hi Jules,
> 
> On Mon, Jul 17, 2023 at 06:54:28PM +0200, Jules Maselbas wrote:
> > Signed-off-by: Jules Maselbas <jmaselbas@kalray.eu>
> > ---
> >  drivers/mtd/spi-nor/Kconfig       |   6 +
> >  drivers/mtd/spi-nor/Makefile      |   1 +
> >  drivers/mtd/spi-nor/dw-ospi-nor.c | 960 ++++++++++++++++++++++++++++++
> >  3 files changed, 967 insertions(+)
> >  create mode 100644 drivers/mtd/spi-nor/dw-ospi-nor.c
> > 
> > diff --git a/drivers/mtd/spi-nor/Kconfig b/drivers/mtd/spi-nor/Kconfig
> > index 2beb26006e..b34c69203e 100644
> > --- a/drivers/mtd/spi-nor/Kconfig
> > +++ b/drivers/mtd/spi-nor/Kconfig
> > @@ -26,4 +26,10 @@ config SPI_CADENCE_QUADSPI
> >  	help
> >  	  This enables support for the Cadence Quad SPI controller and NOR flash.
> >  
> > +config SPI_SYNOPSYS_OCTALSPI_NOR
> > +	tristate "Synopsys DesignWare Octal SPI controller"
> > +	help
> > +	  This enables support for the Synopsys DesignWare Octal SPI controller
> > +	  and NOR flash.
> > +
> >  endif
> > diff --git a/drivers/mtd/spi-nor/Makefile b/drivers/mtd/spi-nor/Makefile
> > index 457a2f0488..61cf789182 100644
> > --- a/drivers/mtd/spi-nor/Makefile
> > +++ b/drivers/mtd/spi-nor/Makefile
> > @@ -1,3 +1,4 @@
> >  # SPDX-License-Identifier: GPL-2.0-only
> >  obj-$(CONFIG_MTD_SPI_NOR)		+= spi-nor.o
> >  obj-$(CONFIG_SPI_CADENCE_QUADSPI)	+= cadence-quadspi.o
> > +obj-$(CONFIG_SPI_SYNOPSYS_OCTALSPI_NOR) += dw-ospi-nor.o
> > diff --git a/drivers/mtd/spi-nor/dw-ospi-nor.c b/drivers/mtd/spi-nor/dw-ospi-nor.c
> > new file mode 100644
> > index 0000000000..13033f9500
> > --- /dev/null
> > +static int dw_spi_find_chipselect(struct spi_nor *nor)
> > +{
> > +	int cs = -1;
> > +	struct dw_spi_nor *dw_spi = nor->priv;
> 
> Better put a pointer to dw_spi_flash_pdata into nor->priv...
> (or use container_of() to get dw_spi_flash_pdata from *nor)

Yeah, I am going to use container_of to get dw_spi_flash_pdata,
and keep dw_spi in priv.
> 
> > +
> > +	for (cs = 0; cs < dw_spi->supported_cs; cs++)
> > +		if (nor == &dw_spi->f_pdata[cs].nor)
> > +			break;
> > +	return cs;
> > +}
> 
> ...and put cs into struct dw_spi_flash_pdata. This way you don't have to
> iterate to get the current chip select.
> 
> > +static int dw_spi_write_reg(struct spi_nor *nor, u8 opcode, u8 *buf, int len)
> > +{
> > +	int i, ret;
> > +
> > +	dev_dbg(nor->dev, "write_reg opcode 0x%02x: ", opcode);
> > +	for (i = 0; i < len; i++)
> > +		pr_debug("%02x ", buf[i]);
> > +	pr_debug("\n");
> > +
> > +	if (!IS_ENABLED(CONFIG_MTD_WRITE))
> > +		return -ENOTSUPP;
> 
> This function looks like it's supposed to write a register. Are you sure
> this is not needed when MTD write support is disabled?

I am not sure about this, this looks exactly the same in the cadence-quadspi
driver. I though adding a condition to only refuse writing to the "write enable"
register, like so:

-       if (!IS_ENABLED(CONFIG_MTD_WRITE))
+       if (!IS_ENABLED(CONFIG_MTD_WRITE) && opcode == SPINOR_OP_WREN)

But I think it should not be handled at the driver level... and it is probably
already handled by core spi-nor driver... In the end I think I'll remove theses
from this driver.

> > +
> > +	ret = dw_spi_prep_std(nor, SPI_TMOD_TO);
> > +	if (ret)
> > +		return ret;
> > +
> > +	return dw_spi_write_std(nor, &opcode, 1, buf, len);
> > +}
> 
> [...]
> 
> > +
> > +	if (!dev->of_node) {
> > +		f_pdata = &dw_spi->f_pdata[0];
> > +
> > +		ret = dw_spi_setup_flash(dev, f_pdata, np);
> > +		if (ret)
> > +			goto probe_failed;
> 
> Do we need probing without device tree? If not, please remove this case.

Only device-tree probing is needed, this likely is a copy-paste from cadence-quadspi
driver.


Best,
Jules