Re: [PATCH 6/6] mtd: spi-nor: core; avoid odd length/address writes in 8D-8D-8D mode

[Pratyush Yadav <p.yadav@xxxxxx>]

On 07/05/21 05:56PM, Michael Walle wrote:
> Am 2021-05-06 21:18, schrieb Pratyush Yadav:
> > On Octal DTR capable flashes like Micron Xcella the writes cannot start
> > or end at an odd address in Octal DTR mode. Extra 0xff bytes need to be
> > appended or prepended to make sure the start address and end address are
> > even. 0xff is used because on NOR flashes a program operation can only
> > flip bits from 1 to 0, not the other way round. 0 to 1 flip needs to
> > happen via erases.
> > 
> > Signed-off-by: Pratyush Yadav <p.yadav@xxxxxx>
> > 
> > ---
> > 
> >  drivers/mtd/spi-nor/core.c | 72 +++++++++++++++++++++++++++++++++++++-
> >  1 file changed, 71 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
> > index 3d66cc34af4d..265d8b25fc7f 100644
> > --- a/drivers/mtd/spi-nor/core.c
> > +++ b/drivers/mtd/spi-nor/core.c
> > @@ -2022,6 +2022,71 @@ static int spi_nor_read(struct mtd_info *mtd,
> > loff_t from, size_t len,
> >  	return ret;
> >  }
> > 
> > +/*
> > + * On Octal DTR capable flashes like Micron Xcella the writes cannot
> > start or
> > + * end at an odd address in Octal DTR mode. Extra 0xff bytes need to
> > be appended
> > + * or prepended to make sure the start address and end address are
> > even. 0xff is
> > + * used because on NOR flashes a program operation can only flip bits
> > from 1 to
> > + * 0, not the other way round. 0 to 1 flip needs to happen via erases.
> > + */
> > +static int spi_nor_octal_dtr_write(struct spi_nor *nor, loff_t to,
> > size_t len,
> > +				   const u8 *buf)
> > +{
> > +	u8 *tmp_buf;
> > +	size_t bytes_written;
> > +	loff_t start, end;
> > +	int ret;
> > +
> > +	if (IS_ALIGNED(to, 2) && IS_ALIGNED(len, 2))
> > +		return spi_nor_write_data(nor, to, len, buf);
> > +
> > +	tmp_buf = kmalloc(nor->page_size, GFP_KERNEL);
> > +	if (!tmp_buf)
> > +		return -ENOMEM;
> > +
> > +	memset(tmp_buf, 0xff, nor->page_size);
> 
> This could be replaced by just setting the first and the
> last byte to 0xff. But this might be easier to read. I am
> fine with both.

First, yes. Not the last. The buffer is allocated to nor->page_size for 
simplicity but the write could be smaller than nor->page_size. So you'd 
need to calculate the position of the other 0xff byte. It is much 
simpler to just initialize the whole buffer. It will be around 256 or 
512 bytes so not a big overhead.

> 
> > +
> > +	start = round_down(to, 2);
> > +	end = round_up(to + len, 2);
> > +
> > +	memcpy(tmp_buf + (to - start), buf, len);
> > +
> > +	ret = spi_nor_write_data(nor, start, end - start, tmp_buf);
> > +	if (ret == 0) {
> > +		ret = -EIO;
> > +		goto out;
> > +	}
> else if ? I've missed this in the other patch.

Following the style used in spi_nor_read(). Anyway, I've seen 
conflicting advice on which style to be used. Some people don't like 
else if when the if ends in a return since it is effectively an else if. 
Others like it the other way round. Dunno...

> 
> > +	if (ret < 0)
> > +		goto out;
> > +
> > +	/*
> > +	 * More bytes are written than actually requested, but that number
> > can't
> > +	 * be reported to the calling function or it will confuse its
> > +	 * calculations. Calculate how many of the _requested_ bytes were
> > +	 * written.
> > +	 */
> > +	bytes_written = ret;
> > +
> > +	if (to != start)
> > +		ret -= to - start;
> > +
> > +	/*
> > +	 * Only account for extra bytes at the end if they were actually
> > +	 * written. For example, if for some reason the controller could only
> > +	 * complete a partial write then the adjustment for the extra bytes at
> > +	 * the end is not needed.
> > +	 */
> > +	if (start + bytes_written == end)
> > +		ret -= end - (to + len);
> > +
> > +	if (ret < 0)
> > +		ret = -EIO;
> 
> can this happen?

I don't think so. IIRC this is left over from when I tried a different 
approach. Maybe I should change it to WARN_ON() to catch future 
programming errors? Though I don't mind if we drop it entirely.

> 
> > +
> > +out:
> > +	kfree(tmp_buf);
> > +	return ret;
> > +}
> > +
> >  /*
> >   * Write an address range to the nor chip.  Data must be written in
> >   * FLASH_PAGESIZE chunks.  The address range may be any size provided
> > @@ -2066,7 +2131,12 @@ static int spi_nor_write(struct mtd_info *mtd,
> > loff_t to, size_t len,
> >  		if (ret)
> >  			goto write_err;
> > 
> > -		ret = spi_nor_write_data(nor, addr, page_remain, buf + i);
> > +		if (nor->write_proto == SNOR_PROTO_8_8_8_DTR)
> > +			ret = spi_nor_octal_dtr_write(nor, addr, page_remain,
> > +						      buf + i);
> > +		else
> > +			ret = spi_nor_write_data(nor, addr, page_remain,
> > +						 buf + i);
> >  		if (ret < 0)
> >  			goto write_err;
> >  		written = ret;
> 
> -michael

Thanks for reviewing.

-- 
Regards,
Pratyush Yadav
Texas Instruments Inc.