UBIFS Panic

[hujianyang]

On 2014/6/27 4:28, Akshay Bhat wrote:
> Hi,
> 
> I was performing a stress test on UBI file system on a ARM based embedded board
> (Beagle Bone). We have a SLC NAND flash and the MTD_UBI_WL_THRESHOLD is set to
> 4096. I have 2 scripts running in the background in a infinite while loop:
> Script1: dd if=/dev/zero of=/var/db/test bs=2M count=1
> Script2: dd if=/dev/urandom of=/var/log/test bs=2M count=1 2> /dev/null
> Note: The above directories are mounted as (sync,relatime).
> 

How did you release data on the flash? What's the partitions on your system?
Did you use MTD_UBI_FASTMAP?

> After the running the scripts for 5 days and max_ec reached the
> MTD_UBI_WL_THRESHOLD. At this point I got panic1 (see below) and the ubi volume
> switched to read-only mode. I rebooted the board, changed the transfer size in
> the script from 2M to 140K and the scripts ran for 2+ days and I got panic 2 (See
> below).

Did you try umount after this error happen and mount partition again, then
re-run your scripts to see what will happen?

> [81438.785011] UBIFS error (pid 31441): do_commit: commit failed, error -30
> [81438.785034] UBIFS error (pid 31441): ubifs_write_inode: can't write inode 79,
> error -30
>

Later error -30 is caused by former error -28 which is reported by UBI layer.
Did you run df to see how much space left on your device?

I think each time you get an error -28 from UBI layer, you will see an ubi_err.
But I didn't see it in your log. Does anyone else know something about it?