JFFS2 Cleanmarker Issue

JFFS2 Cleanmarker Issue

[Suresh Rajashekara]

Hi Everyone,

Few necessary details before I explain my issue.

Hardware I use: OMAP5912 based board
Linux: 2.6.16-rc3

Let me know if you guys need more details.

The issue:
I use the JFFS2 file system on the oneNAND flash. Every time I boot the 
board I see the following on the serial console (I have
CONFIG_JFFS2_FS_DEBUG=2)

<SNIP>
jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block 
at 00000000 OOB at 00000000 was ff ff 00 19 03 03 ff ff ff ff ff 56 fe 
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
ff ff ff jffs_check_nand_cleanmarker returned 1 jffs2_check_oob_empty 
returned 1
JFFS2: Erase block at 0x00000000 is not formatted. It will be erased
jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block 
at 00020000 OOB at 00020000 was ff ff 00 19 03 03 ff ff ff ff ff 56 fe 
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
ff ff ff jffs_check_nand_cleanmarker returned 1 jffs2_check_oob_empty 
returned 1 blah..
.
blah...
.
.
</SNIP>

Getting this on the first boot, is I guess expected. But this messages 
repeats every time we reboot the board. I did some debugging and found 
that the JFFS2 driver is marking these blocks and writes the clean markers.

Since these prints were appearing on every boot, I read the oob area 
which the jffs2 driver wrote to mark the cleanmarkers. I found the data 
to be shifted. I tried Googling and also went through the GIT log 
messages to find if this was a known issue, but could not find any. Is 
this a known issue in 2.6.16-rc3.

The developers who did the initial design have chosen to format the 
partition using the nftl_format command. I read on the MTD howto that it 
should not be used. I tried replacing that with "flash_eraseall -j" and 
created the jffs2 file system using the command mkfs.jffs2 (I tried with 
the correct page size and other stuff by looking in to the datasheet) 
and wrote the jffs2 file system using the nand_write command, but none 
helped.

Could someone please point me in the right direction?

Thanks in advance,

Best Regards,
Suresh

Re: JFFS2 Cleanmarker Issue

[Sudeep K N]

Hi,

The clean marker is not getting written to spare area on your nand device.
The first 4 bytes of clean maker is 19852003 I think.
Just check whether you are getting spare write request for these bytes in your
Nand driver and also check if you have written it properly to the spare area.

On Tue, Jun 17, 2008 at 4:36 AM, Suresh Rajashekara
<suresh@mistralsolutions.com> wrote:
> Hi Everyone,
>
> Few necessary details before I explain my issue.
>
> Hardware I use: OMAP5912 based board
> Linux: 2.6.16-rc3
>
> Let me know if you guys need more details.
>
> The issue:
> I use the JFFS2 file system on the oneNAND flash. Every time I boot the
> board I see the following on the serial console (I have
> CONFIG_JFFS2_FS_DEBUG=2)
>
> <SNIP>
> jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block
> at 00000000 OOB at 00000000 was ff ff 00 19 03 03 ff ff ff ff ff 56 fe
> ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
> ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
> ff ff ff jffs_check_nand_cleanmarker returned 1 jffs2_check_oob_empty
> returned 1
> JFFS2: Erase block at 0x00000000 is not formatted. It will be erased
> jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block
> at 00020000 OOB at 00020000 was ff ff 00 19 03 03 ff ff ff ff ff 56 fe
> ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
> ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
> ff ff ff jffs_check_nand_cleanmarker returned 1 jffs2_check_oob_empty
> returned 1 blah..
> .
> blah...
> .
> .
> </SNIP>
>
> Getting this on the first boot, is I guess expected. But this messages
> repeats every time we reboot the board. I did some debugging and found
> that the JFFS2 driver is marking these blocks and writes the clean markers.
>
> Since these prints were appearing on every boot, I read the oob area
> which the jffs2 driver wrote to mark the cleanmarkers. I found the data
> to be shifted. I tried Googling and also went through the GIT log
> messages to find if this was a known issue, but could not find any. Is
> this a known issue in 2.6.16-rc3.
>
> The developers who did the initial design have chosen to format the
> partition using the nftl_format command. I read on the MTD howto that it
> should not be used. I tried replacing that with "flash_eraseall -j" and
> created the jffs2 file system using the command mkfs.jffs2 (I tried with
> the correct page size and other stuff by looking in to the datasheet)
> and wrote the jffs2 file system using the nand_write command, but none
> helped.
>
> Could someone please point me in the right direction?
>
> Thanks in advance,
>
> Best Regards,
> Suresh
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>



-- 
Regards,
Sudeep

Re: JFFS2 Cleanmarker Issue

[Sudeep K N]

On Wed, Jun 18, 2008 at 12:24 AM, Suresh Rajashekara
<suresh@mistralsolutions.com> wrote:
> Sudeep K N wrote:
>>
>> Hi,
>>
>> The clean marker is not getting written to spare area on your nand device.
>> The first 4 bytes of clean maker is 19852003 I think.
>> Just check whether you are getting spare write request for these bytes in
>> your
>> Nand driver and also check if you have written it properly to the spare
>> area.
>>
>
> Thanks Sudeep. I think the oneNAND driver has some issues. Now I guess I am
> in the right direction.
>
> One more question I have is, why the first 2 bytes are ff ff and the
> cleanmarker is starting from the 3rd byte? Is this how it is? I thought the
> cleanmarker should start from the first byte.
>
> Thanks in advance,
>
> Best Regards,
> Suresh
>

No cleanmarker doesnot always start from 1st byte
The position of cleanmarker depends on the oobfree area specified by
you through
struct nand_ecclayout

-- 
Regards,
Sudeep

Re: JFFS2 Cleanmarker Issue

[suresh]

Hi

Quoting Sudeep K N <sudeepholla.maillist@gmail.com>:
> No cleanmarker doesnot always start from 1st byte
> The position of cleanmarker depends on the oobfree area specified by
> you through
> struct nand_ecclayout
>
> --
> Regards,
> Sudeep

I think I found the reason for the issue. I forgot to mention that we  
are using oneNAND. OneNAND uses the fragmented OOB, which I think is  
not supported by Jffs2 and oneNAND driver in 2.6.16-rc3.

Thanks,
Suresh

Re: JFFS2 Cleanmarker Issue

[Kyungmin Park]

Hi,

On Mon, Jun 30, 2008 at 12:08 AM,  <suresh@mistralsolutions.com> wrote:
> Hi
>
> Quoting Sudeep K N <sudeepholla.maillist@gmail.com>:
>> No cleanmarker doesnot always start from 1st byte
>> The position of cleanmarker depends on the oobfree area specified by
>> you through
>> struct nand_ecclayout
>>
>> --
>> Regards,
>> Sudeep
>
> I think I found the reason for the issue. I forgot to mention that we
> are using oneNAND. OneNAND uses the fragmented OOB, which I think is
> not supported by Jffs2 and oneNAND driver in 2.6.16-rc3.
>
> Thanks,
> Suresh
>

In the early stage, it doesn't handle not-alinged oob access and use
the MTD_OOB_PLACE.
If you don't need to use 2.6.16-rc3, please use the latest kernel or
later version.
or simple hack, please modify oobfree value from {2, 3} to {2, 4},
{18, 3->4} and so one, but not recommend.

Thank you,
Kyungmin Park