Writing from dingux seems to corrupt the FAT32 file system on some cards.

[GoogleCodeExporter]

Need to get hold of one of the failing cards. Any volunteer to donate his?.

Original issue reported on code.google.com by igga...@gmail.com on 30 Jul 2009 at 1:15

[GoogleCodeExporter]

Corruption seems to happen in all cards and during writes.

Setting the MMC interface in 1-bit mode fixes the problem, but clobbers 
throughput.
Released 1-bit mode kernel as a temporary workaround.

Original comment by igga...@gmail.com on 13 Aug 2009 at 2:44

• Added labels: ****
• Removed labels: ****

[GoogleCodeExporter]

good find, so now we could double check kernel code inside #ifdef mmc 1 vs 4 
bit and
search for something suspicious

Ignacio, is there some good way to relatively quickly cause the corruption or 
is it
really random? Is the bad data pure noise or is there some pattern visible 
inside bad
data?  Could you attach example of good and bad data if you already have it?

Did you try to disable other drivers or hardware (LCD) while MMC code writes to 
card
to see if it goes away?

Did you try to compile MMC code as a module? This could be fastest way to try 
changes
- boot from ramdisk, insert mmc module, check for possible corruption, remove,
recompile/reinsert module, try again

Original comment by fano...@gmail.com on 13 Aug 2009 at 8:37

• Added labels: ****
• Removed labels: ****