[PATCH 4/4] mtd: nand: omap2: Add data correction support

[Ivan Djelic]

On Fri, Oct 05, 2012 at 09:51:50AM +0100, Philip, Avinash wrote:
> On Thu, Oct 04, 2012 at 15:51:03, Philip, Avinash wrote:
> > On Thu, Oct 04, 2012 at 00:50:45, Ivan Djelic wrote:
> > > On Wed, Oct 03, 2012 at 03:29:49PM +0100, Philip, Avinash wrote:
> > > > ELM module can be used for error correction of BCH 4 & 8 bit. Also
> > > > support read & write page in one shot by adding custom read_page &
> > > > write_page methods. This helps in optimizing code.
> > > > 
> > > > New structure member "is_elm_used" is added to know the status of
> > > > whether the ELM module is used for error correction or not.
> > > > 
> > > > Note:
> > > > ECC layout of BCH8 uses 14 bytes for 512 byte of data to make compatible
> > > > with RBL ECC layout, even though the requirement was only 13 byte. This
> > > > results a common ecc layout across RBL, U-boot & Linux.
> > > > 
> > > 
> > > See a few comments below,
> > > 
> > > (...)
> > > > +static int omap_elm_correct_data(struct mtd_info *mtd, u_char *dat,
> > > > +                               u_char *read_ecc, u_char *calc_ecc)
> > > > +{
> > > > +       struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
> > > > +                       mtd);
> > > > +       int eccsteps = info->nand.ecc.steps;
> > > > +       int i , j, stat = 0;
> > > > +       int eccsize, eccflag, size;
> > > > +       struct elm_errorvec err_vec[ERROR_VECTOR_MAX];
> > > > +       u_char *ecc_vec = calc_ecc;
> > > > +       enum bch_ecc type;
> > > > +       bool is_error_reported = false;
> > > > +
> > > > +       /* initialize elm error vector to zero */
> > > > +       memset(err_vec, 0, sizeof(err_vec));
> > > > +       if (info->nand.ecc.strength == BCH8_MAX_ERROR) {
> > > > +               size = BCH8_SIZE;
> > > > +               eccsize = BCH8_ECC_OOB_BYTES;
> > > > +               type = BCH8_ECC;
> > > > +       } else {
> > > > +               size = BCH4_SIZE;
> > > > +               eccsize = BCH4_SIZE;
> > > > +               type = BCH4_ECC;
> > > > +       }
> > > > +
> > > > +       for (i = 0; i < eccsteps ; i++) {
> > > > +               eccflag = 0;    /* initialize eccflag */
> > > > +
> > > > +               for (j = 0; (j < eccsize); j++) {
> > > > +                       if (read_ecc[j] != 0xFF) {
> > > > +                               eccflag = 1;    /* data area is flashed */
> > > 
> > > Just a reminder: this way of checking if a page has been programmed is not robust to bitflips,
> > > so you may get into trouble with UBIFS on a fairly recent device.
> 
> Sorry I missed this point.
> 
> Here we were checking data in spare area (only in ecc layout 14*4 for BCH8) is 0xFF. If all data
> in spare area is 0xFF, conclude that the page is erased & no need of error correction. In case
> of bit flip present in spare area, page will be reported as uncorrectable.
> But I am not understand understand " trouble with UBIFS on a fairly recent device".
> Can you little elaborativ

There are at least 2 potential problems when reading an erased page with bitflips:

1. bitflip in data area and no bitflip in spare area (all 0xff)
Your code will not perform any ECC correction.
UBIFS does not like finding bitflips in empty pages, see for instance
http://lists.infradead.org/pipermail/linux-mtd/2012-March/040328.html.

2. bitflip in ECC bytes in spare area
Your code will report an uncorrectable error upon reading; if this happens while reading a partially programmed UBI block,
I guess you will lose data.

BR,
--
Ivan