[RESUBMIT] [PATCH] [MTD] NAND nand_ecc.c: rewrite for improved performance

Frans Meulenbroeks fransmeulenbroeks at gmail.com
Mon Aug 18 02:40:28 EDT 2008


2008/8/18, Troy Kisky <troy.kisky at boundarydevices.com>:
> > +     if (nr_bits == 11) {    /* correctable error */
>
>
> This is a necessary, but NOT sufficient condition to
>  determine that it is a 1 bit error.

The test for 11 bits is in accordance to the ST datasheet I used
(http://www.st.com/stonline/books/pdf/docs/10123.pdf, see section
3.4).
What other check do you feel should be needed.
>
>
>  Big MAYBE below
>
>  This is probably why you passed the big endian test.
>  It was seeing every read as a single bit error
>  and silently correcting. Adding a debug print
>  when correcting or failing to correct would be
>  useful.

As long as they are debug print, that could indeed be done. No problem
with that.
We could even have a printk to indicate a failure on the console/in the log.
Having a printk when a correction is done is probably not that
desirable (newer generations of nand seem more and more often to
require corrections, which could mean too many msges in the log).

Frans.



More information about the linux-mtd mailing list