CFI numblocks incorrect?
it seems CFI flash support is somehow broken. mtd->eraseregions is filled as bellow:
erase size blocks correct blocks 00004000 01 01 00002000 01 02 00008000 00 01 00010000 0f 0f
'blocks' is actual numblocks field value while 'correct blocks' is expected value. flinfo gives correct results (it is AM29LV800BB). It seems to be related to cfi_mtd_set_erasesize function, but I'm too tired to debug it further. Just reporting in case anyone has time to look at it. Otherwise I'll do so at Friday.
Best regards,
ladis
PATCH 1/1 CFI: fix eraseregions numblocks
From: Ladislav Michl <[email protected]>
eraseregions numblocks was sometimes one less than actual, possibly producing erase regions with zero blocks. As MTD code touches eraseregions only if numeraseregions is greater that zero, allocate eraseregions only for non uniform erase size flash.
--- cfi_mtd.c | 35 ++++++++++++++++++----------------- 1 file changed, 18 insertions(+), 17 deletions(-)
drivers/mtd/cfi_mtd.c +18 -17
@@ -161,8 +161,8 @@ static int cfi_mtd_set_erasesize(struct mtd_info *mtd, flash_info_t *fi)int sect;int regions =0 ;int numblocks =0 ; -ulong offset =0 ; -ulong base_addr = fi->start[0 ]; +ulong offset; +ulong base_addr;/* * First detect the number of eraseregions so that we can allocate
@@ -174,29 +174,35 @@ static int cfi_mtd_set_erasesize(struct mtd_info *mtd, flash_info_t *fi) sect_size_old =flash_sector_size (fi, sect); } +switch (regions) { +case 0 : +return 1 ; +case 1 :/* flash has uniform erase size */ + mtd->numeraseregions =0 ; + mtd->erasesize = sect_size_old; +return 0 ; + } + + mtd->numeraseregions = regions; mtd->eraseregions =malloc (sizeof (struct mtd_erase_region_info ) * regions);/* * Now detect the largest sector and fill the eraseregions */ - sect_size_old =0 ; regions =0 ; + base_addr = offset = fi->start[0 ]; + sect_size_old =flash_sector_size (fi,0 );for (sect =0 ; sect < fi->sector_count; sect++) { -if ((sect_size_old !=flash_sector_size (fi, sect)) && - (sect_size_old !=0 )) { +if (sect_size_old !=flash_sector_size (fi, sect)) { mtd->eraseregions[regions].offset = offset - base_addr; mtd->eraseregions[regions].erasesize = sect_size_old; mtd->eraseregions[regions].numblocks = numblocks; -/* Now start counting the next eraseregions */ numblocks =0 ; regions++; - }else { - numblocks++; - } - -if (sect_size_old !=flash_sector_size (fi, sect)) offset = fi->start[sect]; + } + numblocks++;/* * Select the largest sector size as erasesize (e.g. for UBI)
@@ -212,12 +218,7 @@ static int cfi_mtd_set_erasesize(struct mtd_info *mtd, flash_info_t *fi)*/ mtd->eraseregions[regions].offset = offset - base_addr; mtd->eraseregions[regions].erasesize = sect_size_old; - mtd->eraseregions[regions].numblocks = numblocks +1 ; - -if (regions) - mtd->numeraseregions = regions +1 ; -else - mtd->numeraseregions =0 ; + mtd->eraseregions[regions].numblocks = numblocks; mtd->erasesize = sect_size;
PATCH 1/1 CFI: fix eraseregions numblocks
Quoted 6 lines · On Thursday 28 January 2010 12:27:14 Ladislav Michl wrote:
From: Ladislav Michl <[email protected]>
eraseregions numblocks was sometimes one less than actual, possibly
producing erase regions with zero blocks. As MTD code touches eraseregions only if numeraseregions is greater that zero, allocate eraseregions only for non uniform erase size flash.
Applied to u-boot-cfi-flash/master. Thanks.
Cheers,
Stefan
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-0 Fax: (+49)-8142-66989-80 Email: office at denx.de