Confuse with WL_THRESHOLD

[Thang Dao Duy]

On Mon, Jun 8, 2009 at 2:26 PM, Artem Bityutskiy<dedekind at infradead.org> wrote:
> On Sun, 2009-06-07 at 13:23 +0300, Artem Bityutskiy wrote:
>> On Sat, 2009-06-06 at 07:03 +0700, Thang Dao Duy wrote:
>> > Hi everyone.
>> > I used NANDsim and ubifs not for along time. Now, I have confused with
>> > wear-leveling threshold. If i give parameter WL_THRESHOLD in Kernel is
>> > small (example, the value is  3), then when i write a file size 2 MiB
>> > to NAND about 200 times, it makes me cannot read EC and VID headers of
>> > one PEB (example PEB 10). May be that PEB is bad ? And when i set the
>> > WL_THRESHOLD is higher (the value is 20), i write a file 2MiB about
>> > 1000 times, nothing 's happen, i can read headers of any PEB. So can
>> > you tell me what make this happen, and show me where in ubi source in
>> > kernel.
>>
>> Looks like a bug. Please, describe precisely how to reproduce it,
>> and I'll take a look at this.
>
> Please, also make sure you are using the latest UBI/UBIFS. Please, take
> a look here:
> http://www.linux-mtd.infradead.org/doc/ubifs.html#L_how_send_bugreport
>
> I've configured my UBI and made the threshold to be 3 now. I've also
> enabled additional UBI and UBIFS debugging checks. I'm running the
> following test:
>
> $ sudo modprobe nandsim
> $ sudo modprobe ubi mtd=0
> $ sudo ubimkvol /dev/ubi0 -m -N test_fs
> Set volume size to 128674304
> Volume ID 0, size 8107 LEBs (128674304 bytes, 122.7 MiB), LEB size 15872 bytes (15.5 KiB), dynamic, name "test_fs", alignment 1
> $ sudo modprobe ubifs
> $ sudo mount -t ubifs ubi0:test_fs /mnt/ubifs/
> $ while true; do sudo dd if=/dev/urandom of=ubifs/file count=2048; sync; done
> 2097152 bytes (2.1 MB) copied, 0.316768 s, 6.6 MB/s
> 4096+0 records in
> 4096+0 records out
> ....
>
> It's being run for some time now, so far so good - no issues.
>
> --
> Best regards,
> Artem Bityutskiy (Битюцкий Артём)
>
>

Thanks to Atem Bityutskiy.
I have modified some code in kernel drivers/mtd/nand/, so may be it
makes that happen. I tested  the useful nand (512 and 2048 byte page
size), it have the same with small or big WL_THRESHOLD. So that is my
fault.