[PATCH v7 0/6] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes

Pekon Gupta pekon at ti.com
Fri Jan 17 02:13:11 EST 2014


*changes v6 -> v7*
[PATCH 1/6] <no change>
[PATCH 2/6] <no change>
[PATCH 3/6] <dropped older version, refer commit message>
 - erased-page detection done based on count of zero-bits in OOB & DATA
 - incorporated some feedbacks from Brian Norris <computersforpeace at gmail.com>
	http://lists.infradead.org/pipermail/linux-mtd/2014-January/051533.html
	http://lists.infradead.org/pipermail/linux-mtd/2014-January/051534.html
[PATCH 4/6] <new> remove redundant bit-flip counting for erased-page
[PATCH 5/6] <minor cleanup>
[PATCH 6/6] <minor cleanup>
(As there are functional changes in this version, so not attaching
 Tested-by: Stefan Roese <sr at denx.de>)


*changes v5 -> v6*
[PATCH 1/6] <no change>
[PATCH 2/6] introduced variable 'actual_eccbytes' to omit reserved byte-position
[PATCH 3/6] split from [PATCH v5 3/5] fix erased-page bit-flip correction
[PATCH 4/6] split from [PATCH v5 3/5] fix erased-page detection
[PATCH 5/6] <minor cleanup>
[PATCH 6/6] <no change>


*changes v4 -> v5*
This patch series was split version of earlier patch:
http://lists.infradead.org/pipermail/linux-mtd/2013-November/050241.html

chip->ecc.correct() is used for detecting and correcting bit-flips during read
operations. In OMAP NAND driver different ecc-schemes have different callbacks:
 - omap_correct_data()		for HAM1_HW ecc-schemes (Untouched)
 - nand_bch_correct_data()	for BCHx_HW_DETECTION_SW ecc-schemes (Untouched)
 - omap_elm_correct_data()	for BCHx_HW ecc-schemes

This patch-series fixes following issues in omap_elm_correct_data():
(1) Dependency on a specific reserved byte-position in OOB area
    to differentiates between erased-pages v/s programmed-pages.
    Problem: reserved byte-position cannot be accomodated in all ecc-schemes
    Problem: reserved byte-position can itself be subjected upto 8 bit-flips
             causing the 0xff to become 0x00, causing page to be
             mis-recognized as erased-page.

(2) Bit-flips in erased-pages are detected by comparing each byte of Data & OOB
    with 0xff in check_erased_page().
    Problem: This is causes performance penalty when erased-pages are checked.

(3) Current code is not scalable for future ECC schemes due to presence of 
    tweaks for BCH4_ECC and BCH8_ECC at multiple places.

(4) Currently, bit-flips are evaluated and fixed even when ELM reports them as
    un-correctable bit-flips, this should not happen as 'number-of-error' field
    in ELM_LOCATION_STATUS becomes invalid when un-correctable flag is set.

(5) Driver should return with error-code = '-EBADMSG' when
     uncorrectable bit-flip is detected
     bit-flip outside valid Data and OOB region is detected


Pekon Gupta (6):
  mtd: nand: omap: add field to indicate current ecc-scheme in 'struct
    omap_nand_info'
  mtd: nand: omap: ecc.correct: omap_elm_correct_data: rename ambiguous
    variable 'eccsize' and 'ecc_vector_size'
  mtd: nand: omap: ecc.correct: omap_elm_correct_data: fix erased-page
    detection for BCHx_HW ECC schemes
  mtd: nand: omap: ecc.correct: omap_elm_correct_data: remove redundant
    bit-flip counting for erased-page
  mtd: nand: omap: ecc.correct: omap_elm_correct_data: cleanup for
    future enhancements
  mtd: nand: omap: ecc.correct: omap_elm_correct_data: fix
    programmed-page bit-flip correction logic

 drivers/mtd/nand/omap2.c | 265 +++++++++++++++++++++++------------------------
 1 file changed, 127 insertions(+), 138 deletions(-)

-- 
1.8.1




More information about the linux-mtd mailing list