write error while using UBI/UBIFS

Manoj manoj23 at gmail.com
Mon Jun 8 22:13:33 EDT 2009

Thanks for the pointers. But I am using kernel v2.6.27.7. :-(

Meanwhile, I disabled all extra checks and
CONFIG_MTD_NAND_VERIFY_WRITE as recommended. Now, I see this when I
unmount and re-mount ubifs:

# mount -o sync -t ubifs ubi0:rootfs /tmp/flash/
[  101.480000] UBIFS error (pid 164): ubifs_scan: corrupt empty space
at LEB 1:14336
[  101.490000] UBIFS error (pid 164): ubifs_scanned_corruption:
corrupted data at LEB 1:14336
[  101.500000] UBIFS error (pid 164): ubifs_scan: LEB 1 scanning failed
[  101.530000] UBIFS error (pid 164): ubifs_recover_master_node:
failed to recover master node
mount: mounting ubi0:rootfs on /tmp/flash/ failed: Invalid argument

On Sun, Jun 7, 2009 at 10:59 PM, Artem Bityutskiy<dedekind at infradead.org> wrote:
> Hi,
> On Fri, 2009-06-05 at 20:05 -0700, Manoj wrote:
>> I have created a ubi image that I have programmed to NAND flash using
>> ubiformat. After I mount the ubifs, I am trying to write a 3MB file to
>> the flash and I am seeing some issues.
> Please, make sure you have all the latest UBI and UBIFS fixes:
> http://www.linux-mtd.infradead.org/doc/ubifs.html#L_source
> I think I fixed debugging problems recently. I suspect this fix may
> help you:
> http://git.infradead.org/ubi-2.6.git?a=commit;h=f9fd3ec55e099ccd6894d7e639e2419c53a823e0
>> I have enabled UBI debug and CONFIG_MTD_NAND_VERIFY_WRITE.
> Please, also take this into account:
> http://www.linux-mtd.infradead.org/faq/ubi.html#L_subpage_verify_fail
>> Also, as an experiment, I used mtd_debug utility to write a 10MB image
>> to the flash and I don't see any issues there. So, it doesn't sound
>> like a driver issue?
>> HW ECC is enabled and I am presuming that no maintenance data is
>> stored in spare area for UBI/UBIFS (from documentation). I don't see
>> errors with ECC off.
> Enabling debugging checks for verification is good. But please, bear
> in mind that they make your system slower.
> If you enable just debugging - it is light-weight, but if you also
> enable additional checks - they are very slow, so you should later
> disable them.
> --
> Best regards,
> Artem Bityutskiy (Битюцкий Артём)

More information about the linux-mtd mailing list