flash_eraseall -j and refusing to mount jffs2 on NAND flash partition

flash_eraseall -j and refusing to mount jffs2 on NAND flash partition

[Deven Balani]

Hi all,

I'm using linux kernel 2.6.28.9 and flash_eraseall -j from mtd-utils-1.2.0 to install jffs2 cleanmarker on a NAND flash partition. 

Doing so I ran into two problems?

1. When I attempted to install jffs2 cleanmarker using flash_eraseall -j, I got following error -

flash_eraseall:/dev/mtd6: unable to get NAND oobinfo

I did some debugging to find that MEMGETOOBSEL ioctl is returning -ENOTTY in flash_eraseall.c. Debugged further to find that there is a size mismatch between "struct nand_oobinfo" defined in mtd-utils-1.2.0/include/mtd/mtd-abi.h 

struct nand_oobinfo {
   uint32_t useecc;
   uint32_t eccbytes;
   uint32_t oobfree[8][2];
   uint32_t eccpos[32];
};

Whereas in linux-2.6.28.9/include/mtd/mtd-abi.h it is defined as -
struct nand_oobinfo {
   uint32_t useecc;
   uint32_t eccbytes;
   uint32_t oobfree[8][2];
   uint32_t eccpos[64];
};

I was able to get around this problem by changing the size of eccpos[] to 64 in mtd-utils-1.2.0/include/mtd/mtd-abi.h.

2. Then I tried flash_eraseall -j /dev/mtd6 on NAND partition, it completed successfully but subsequent mount -t jffs2 failed -

# mount -t jffs2 /dev/mtdblock6 /mnt
Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes
empty_blocks 0, bad_blocks 0, c->nr_blocks 320
mount: Mounting /dev/mtdblock6 on /mnt failed: Input/output error

I google'd and found that there has been some discussion on this long back and following was the last post.

> Re: Flash_eraseall -j and cowardly refusing to mount
>
> Jeff Hane jhane at mobilygen.com
> Mon Jan 7 13:39:19 EST 2008
>
> But doesn't this only fix the case where you have contiguous 8 bytes to
> write the cleanmarker?  flash_eraseall -j uses MTD_PLACE to write the
> cleanmarker BUT jffs2 uses MTD_AUTO to read the cleanmarker.
> For example, if your oob area has byte 0 as the bad block indicator and
> bytes 8-10 have ecc, there are only 7 consecutive free bytes.
> flash_eraseall -j will only write 7 bytes so jffs2 will fail when
> mounting the filesystem because it will read 8 bytes.  The 8th byte,
> which will be byte 12, will still be FF.
>  It seems there should be an ioctl that would allow one to read/write
> the oob area using MTD_AUTO???

> jeff

I still see that mtd-utils still don't support this feature. Are there any plans to support this feature in mtd-utils?

I can get around this problem, if I do just flash_eraseall without -j option and then using "mount -t jffs2 -rw" so that jffs2 cleanmarkers are installed during mount time.

I'm not blocked by these problems anyways just thought to bring it to the notice of community. 

Comments and suggestions are welcome.

Thanks,
Deven

Re: flash_eraseall -j and refusing to mount jffs2 on NAND flash partition

[stanley.miao]

I encountered the same problem.
I wrote several patch to mtd-utils and I will post them here later.

Stanley.

Deven Balani wrote:
> Hi all,
>
> I'm using linux kernel 2.6.28.9 and flash_eraseall -j from mtd-utils-1.2.0 to install jffs2 cleanmarker on a NAND flash partition. 
>
> Doing so I ran into two problems?
>
> 1. When I attempted to install jffs2 cleanmarker using flash_eraseall -j, I got following error -
>
> flash_eraseall:/dev/mtd6: unable to get NAND oobinfo
>
> I did some debugging to find that MEMGETOOBSEL ioctl is returning -ENOTTY in flash_eraseall.c. Debugged further to find that there is a size mismatch between "struct nand_oobinfo" defined in mtd-utils-1.2.0/include/mtd/mtd-abi.h 
>
> struct nand_oobinfo {
>    uint32_t useecc;
>    uint32_t eccbytes;
>    uint32_t oobfree[8][2];
>    uint32_t eccpos[32];
> };
>
> Whereas in linux-2.6.28.9/include/mtd/mtd-abi.h it is defined as -
> struct nand_oobinfo {
>    uint32_t useecc;
>    uint32_t eccbytes;
>    uint32_t oobfree[8][2];
>    uint32_t eccpos[64];
> };
>
> I was able to get around this problem by changing the size of eccpos[] to 64 in mtd-utils-1.2.0/include/mtd/mtd-abi.h.
>
> 2. Then I tried flash_eraseall -j /dev/mtd6 on NAND partition, it completed successfully but subsequent mount -t jffs2 failed -
>
> # mount -t jffs2 /dev/mtdblock6 /mnt
> Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes
> empty_blocks 0, bad_blocks 0, c->nr_blocks 320
> mount: Mounting /dev/mtdblock6 on /mnt failed: Input/output error
>
> I google'd and found that there has been some discussion on this long back and following was the last post.
>
>   
>> Re: Flash_eraseall -j and cowardly refusing to mount
>>
>> Jeff Hane jhane at mobilygen.com
>> Mon Jan 7 13:39:19 EST 2008
>>
>> But doesn't this only fix the case where you have contiguous 8 bytes to
>> write the cleanmarker?  flash_eraseall -j uses MTD_PLACE to write the
>> cleanmarker BUT jffs2 uses MTD_AUTO to read the cleanmarker.
>> For example, if your oob area has byte 0 as the bad block indicator and
>> bytes 8-10 have ecc, there are only 7 consecutive free bytes.
>> flash_eraseall -j will only write 7 bytes so jffs2 will fail when
>> mounting the filesystem because it will read 8 bytes.  The 8th byte,
>> which will be byte 12, will still be FF.
>>  It seems there should be an ioctl that would allow one to read/write
>> the oob area using MTD_AUTO???
>>     
>
>   
>> jeff
>>     
>
> I still see that mtd-utils still don't support this feature. Are there any plans to support this feature in mtd-utils?
>
> I can get around this problem, if I do just flash_eraseall without -j option and then using "mount -t jffs2 -rw" so that jffs2 cleanmarkers are installed during mount time.
>
> I'm not blocked by these problems anyways just thought to bring it to the notice of community. 
>
> Comments and suggestions are welcome.
>
> Thanks,
> Deven
>
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>
>