Re: [PATCH v4 3/5] mtd: rawnand: meson: always read whole OOB bytes

[Miquel Raynal]

Hi Arseniy,

avkrasnov@xxxxxxxxxxxxxx wrote on Tue, 23 May 2023 20:27:35 +0300:

> On 22.05.2023 18:38, Miquel Raynal wrote:
> > Hi Arseniy,
> > 
> > AVKrasnov@xxxxxxxxxxxxxx wrote on Mon, 15 May 2023 12:44:37 +0300:
> >   
> >> This changes size of read access to OOB area by reading all bytes of
> >> OOB (free bytes + ECC engine bytes).  
> > 
> > This is normally up to the user (user in your case == jffs2). The
> > controller driver should expose a number of user accessible bytes and
> > then when users want the OOB area, they should access it entirely. On
> > top of that read, they can extract (or "write only") the user bytes.  
> 
> Sorry, I didn't get it. If driver exposes N bytes of user accessible bytes,
> I must always return whole OOB yes? E.g. N + rest of OOB

Yes. At the NAND controller level, you get asked for either a page of
data (sometimes a subpage, but whatever), and/or the oob area. You need
to provide what is requested, no more, no less. The upper layers will
trim down what's uneeded and extract the bytes they want.

> >> Signed-off-by: Arseniy Krasnov <AVKrasnov@xxxxxxxxxxxxxx>
> >> ---
> >>  drivers/mtd/nand/raw/meson_nand.c | 24 ++++++++++++++++++++++++
> >>  1 file changed, 24 insertions(+)
> >>
> >> diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c
> >> index 8526a6b87720..a31106c943d7 100644
> >> --- a/drivers/mtd/nand/raw/meson_nand.c
> >> +++ b/drivers/mtd/nand/raw/meson_nand.c
> >> @@ -755,6 +755,30 @@ static int __meson_nfc_read_oob(struct nand_chip *nand, int page,
> >>  	u32 oob_bytes;
> >>  	u32 page_size;
> >>  	int ret;
> >> +	int i;
> >> +
> >> +	/* Read ECC codes and user bytes. */
> >> +	for (i = 0; i < nand->ecc.steps; i++) {
> >> +		u32 ecc_offs = nand->ecc.size * (i + 1) +
> >> +			       NFC_OOB_PER_ECC(nand) * i;
> >> +
> >> +		ret = nand_read_page_op(nand, page, 0, NULL, 0);
> >> +		if (ret)
> >> +			return ret;
> >> +
> >> +		/* Use temporary buffer, because 'nand_change_read_column_op()'
> >> +		 * seems work with some alignment, so we can't read data to
> >> +		 * 'oob_buf' directly.  
> > 
> > DMA?  
> 
> Yes I guess, this address passed to exec_op code and used as DMA.

If your controller uses DMA on exec_op accesses, then yes. Exec_op
reads/writes are usually small enough (or not time sensitive at all if
they are bigger) so it's not required to use DMA there. Anyhow, oob_buf
is suitable for DMA purposes, so I'm a bit surprised you need a bounce
buffer, if that's the only reason. Maybe you need a bounce buffer to
reorganize the data. That would be a much better explanation.

> >> +		 */
> >> +		ret = nand_change_read_column_op(nand, ecc_offs, meson_chip->oob_buf,
> >> +						 NFC_OOB_PER_ECC(nand), false);
> >> +		if (ret)
> >> +			return ret;
> >> +
> >> +		memcpy(oob_buf + i * NFC_OOB_PER_ECC(nand),
> >> +		       meson_chip->oob_buf,
> >> +		       NFC_OOB_PER_ECC(nand));
> >> +	}
> >>  
> >>  	oob_bytes = meson_nfc_get_oob_bytes(nand);
> >>    
> > 
> > 
> > Thanks,
> > Miquèl  
> 
> Thanks, Arseniy


Thanks,
Miquèl