PXA255 with Samsung K9K1G08UOM flash problem

Kokachev kokachev at kvant-efir.com.ua
Fri Jun 24 05:00:04 EDT 2005


I have Xscale PXA255 SBC with Samsung K9K1G08UOM flash on-board. I
wrote NAND driver for this flash. I use kernel 2.6.7 and MTD snapshot
from 20 of June 2005. And when I make insmod of my driver I see a lot
of messages: "Bad eraseblock 0 from 00000000","Bad eraseblock 1 from
00004000",etc. I began to read the source and I see that from OOB
blocks of my flash I read non-FF values. When I inserted some printk
functions to see what I read from OOB blocks the count of bad
eraseblocks decreased. My flash is connected at address 0x04000000 of
Xscale memory map. ALE and CLE are connected to GPIOs. I think, that
this could be cache problem. I wrote my own nand_write_byte function
with Drain (&Fill) cache command after writing the byte. Situation
changed. function nand_read_oob function reads right values, but
bbt_scan takes about 30 minutes!!! Even without "bad eraseblock" messages when
I try to make nandwrite I receive the message "File cannot fit in to
media due to bad blocks". What should I do? Is it possible to saw to
Xscale, that memory at address 0x04000000 should not be cacheable?






More information about the linux-mtd mailing list