Letux Kernel

Issue 528: 3.13: UBIFS/NAND errors

Reported by Nikolaus Schaller, Dec 8, 2013

NAND config has changed ECC to HW and therefore prints tons of 
errors when mounting 3.12 NAND image on a device where 
NAND&UBIFS is automounted by /etc/fstab (like the hw-validation 
Debian Wheezy).

This delays the boot process by approx. 10-30 minutes.

The issue can be solved by erasing the NAND partition through 
U-Boot. But only a minority of users has access to RS232, so they 
are not able to to it at all.

And people who just want to try a 3.13 image on SD but who want to 
keep the NAND image intact are excluded by the new kernel.

=> driver should ignore such errors after 5 or 10 attempts (or 
report them more severely so that the UBIFS layer does no longer try 
to mount)

Anyways, a fresh 3.13 compatible UBIFS needs to clear/reinitialize 
NAND by

flash_eraseall -j /dev/mtd4

after booting.

Comment 1 by Nikolaus Schaller, Jan 6, 2014

we have hacked a change in the ubifs io system so that it returns an 
-EIO after 3 failed attempts to read NAND.

This is neither a good solution nor a long-term solution, but allows 
to boot a 3.13 kernel with UBIFS automounting enabled but a 3.12 
compatible file system in NAND.

Commit: 74b3680de591221a3490bee049dd8911c5f91b8d
Status: Fixed

Created: 10 years 6 months ago by Nikolaus Schaller

Updated: 10 years 5 months ago

Status: Fixed

Labels:
Type:Defect
Priority:Critical
Milestone:3.13