The logic wants 512-byte aligned blk ops.
To switch to byte-based block accesses, the fixed commit changed the
blk read offset,
PAGE_START(addr) >> 9
with
PAGE_START(addr)
which min alignment, for on-drive OOB, is the min OOB size.
Consequently the reads are offset by PAGE_START(addr) & 0x1ff.
Fixes: 9fc0d361cc41 ("nand: Switch to byte-based block access")
Cc: Eric Blake <ebl...@redhat.com>
Signed-off-by: Karl Beldan <karl.beldan+...@gmail.com>
---
hw/block/nand.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/hw/block/nand.c b/hw/block/nand.c
index 919cb9b803..ed587f60f0 100644
--- a/hw/block/nand.c
+++ b/hw/block/nand.c
@@ -788,7 +788,7 @@ static void glue(nand_blk_load_, PAGE_SIZE)(NANDFlashState
*s,
OOB_SIZE);
s->ioaddr = s->io + SECTOR_OFFSET(s->addr) + offset;
} else {
- if (blk_pread(s->blk, PAGE_START(addr), s->io,
+ if (blk_pread(s->blk, PAGE_START(addr) & ~0x1ff, s->io,
(PAGE_SECTORS + 2) << BDRV_SECTOR_BITS) < 0) {
printf("%s: read error in sector %" PRIu64 "\n",
__func__, PAGE_START(addr) >> 9);
--
2.16.1.72.g5be1f00
