FAT file system on NAND flash memory

FAT file system on NAND flash memory

[Seongsu Lee]

Hi,

In my knowledge, mtdblock.c must be used to use 
FAT filesystem on bare NAND flash memories 
because NFTL is made only for a specific hardware, 
DiskOnChip.

(Layer 0) VFS
(Layer 1) FAT (or EXT2, EXT3)
(Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
(Layer 3) NAND flash memory driver (vendor specific)
(Layer 4) Physical NAND flash memory

Please let me know if it is wrong.

-- 
Seongsu's personal blog - http://www.senux.com/
LILO, you've got me on my knees! (from David Black,
dblack@pilot.njin.net, with apologies to Derek and the
Dominos, and Werner Almsberger)

Re: FAT file system on NAND flash memory

[Josh Boyer]

On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
> Hi,
>
> In my knowledge, mtdblock.c must be used to use
> FAT filesystem on bare NAND flash memories
> because NFTL is made only for a specific hardware,
> DiskOnChip.
>
> (Layer 0) VFS
> (Layer 1) FAT (or EXT2, EXT3)
> (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
> (Layer 3) NAND flash memory driver (vendor specific)
> (Layer 4) Physical NAND flash memory
>
> Please let me know if it is wrong.

That layering is correct, but your results will be bad.  mtdblock
doesn't handle bad blocks.  So if you have one in the middle of your
filesystem, it will break things.  You need some kind of translation
layer if you're going to use NAND.

josh

Re: FAT file system on NAND flash memory

[Seongsu Lee]

On Thu, Apr 12, 2007 at 08:55:23AM -0500, Josh Boyer wrote:
> On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
> > Hi,
> >
> > In my knowledge, mtdblock.c must be used to use
> > FAT filesystem on bare NAND flash memories
> > because NFTL is made only for a specific hardware,
> > DiskOnChip.
> >
> > (Layer 0) VFS
> > (Layer 1) FAT (or EXT2, EXT3)
> > (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
> > (Layer 3) NAND flash memory driver (vendor specific)
> > (Layer 4) Physical NAND flash memory
> >
> > Please let me know if it is wrong.
> 
> That layering is correct, but your results will be bad.  mtdblock
> doesn't handle bad blocks.  So if you have one in the middle of your
> filesystem, it will break things.  You need some kind of translation
> layer if you're going to use NAND.

Thank you for the reply. 

Just one more question.

Could you tell me the reason that there is no translation layer code
in MTD for bare flash memories? 
(Because of patent or copyright? or technical reason? I wonder.)

-- 
Seongsu's personal blog - http://www.senux.com/
Beauty may be skin deep, but ugly goes clear to the
bone. -- Redd Foxx

Re: FAT file system on NAND flash memory

[Matthieu CASTET]

Hi,

Josh Boyer a écrit :
> On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
>> Hi,
>>
>> In my knowledge, mtdblock.c must be used to use
>> FAT filesystem on bare NAND flash memories
>> because NFTL is made only for a specific hardware,
>> DiskOnChip.
>>
>> (Layer 0) VFS
>> (Layer 1) FAT (or EXT2, EXT3)
>> (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
>> (Layer 3) NAND flash memory driver (vendor specific)
>> (Layer 4) Physical NAND flash memory
>>
>> Please let me know if it is wrong.
> 
> That layering is correct, but your results will be bad.  mtdblock
> doesn't handle bad blocks.  So if you have one in the middle of your
> filesystem, it will break things.  You need some kind of translation
> layer if you're going to use NAND.
> 
But UBI could be used between nand and mtdblock ?


Matthieu

Re: FAT file system on NAND flash memory

[Josh Boyer]

On 4/26/07, Seongsu Lee <senux@senux.com> wrote:
> On Thu, Apr 12, 2007 at 08:55:23AM -0500, Josh Boyer wrote:
> > On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
> > > Hi,
> > >
> > > In my knowledge, mtdblock.c must be used to use
> > > FAT filesystem on bare NAND flash memories
> > > because NFTL is made only for a specific hardware,
> > > DiskOnChip.
> > >
> > > (Layer 0) VFS
> > > (Layer 1) FAT (or EXT2, EXT3)
> > > (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
> > > (Layer 3) NAND flash memory driver (vendor specific)
> > > (Layer 4) Physical NAND flash memory
> > >
> > > Please let me know if it is wrong.
> >
> > That layering is correct, but your results will be bad.  mtdblock
> > doesn't handle bad blocks.  So if you have one in the middle of your
> > filesystem, it will break things.  You need some kind of translation
> > layer if you're going to use NAND.
>
> Thank you for the reply.
>
> Just one more question.
>
> Could you tell me the reason that there is no translation layer code
> in MTD for bare flash memories?

Because most users of MTD already handle bad blocks themselves on bare
NAND.  JFFS2, yaffs, etc. all have bad block handling contained within
them.

UBI can be used for things that lack the proper translation
themselves.  JFFS2 can even run on top if a UBI volume.  Perhaps you
should check into that.

josh

Re: FAT file system on NAND flash memory

[Josh Boyer]

On 4/27/07, Matthieu CASTET <matthieu.castet@parrot.fr> wrote:
> Hi,
>
> Josh Boyer a écrit :
> > On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
> >> Hi,
> >>
> >> In my knowledge, mtdblock.c must be used to use
> >> FAT filesystem on bare NAND flash memories
> >> because NFTL is made only for a specific hardware,
> >> DiskOnChip.
> >>
> >> (Layer 0) VFS
> >> (Layer 1) FAT (or EXT2, EXT3)
> >> (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
> >> (Layer 3) NAND flash memory driver (vendor specific)
> >> (Layer 4) Physical NAND flash memory
> >>
> >> Please let me know if it is wrong.
> >
> > That layering is correct, but your results will be bad.  mtdblock
> > doesn't handle bad blocks.  So if you have one in the middle of your
> > filesystem, it will break things.  You need some kind of translation
> > layer if you're going to use NAND.
> >
> But UBI could be used between nand and mtdblock ?

Yep, indeed it can.

josh

Re: FAT file system on NAND flash memory

[Seongsu Lee]

On Thu, Apr 12, 2007 at 08:55:23AM -0500, Josh Boyer wrote:
> On 4/12/07, Seongsu Lee <senux@senux.com> wrote:
> > Hi,
> >
> > In my knowledge, mtdblock.c must be used to use
> > FAT filesystem on bare NAND flash memories
> > because NFTL is made only for a specific hardware,
> > DiskOnChip.
> >
> > (Layer 0) VFS
> > (Layer 1) FAT (or EXT2, EXT3)
> > (Layer 2) mtdblock.c (nftl.c can be used only with DiskOnChip)
> > (Layer 3) NAND flash memory driver (vendor specific)
> > (Layer 4) Physical NAND flash memory
> >
> > Please let me know if it is wrong.
> 
> That layering is correct, but your results will be bad.  mtdblock
> doesn't handle bad blocks.  So if you have one in the middle of your
> filesystem, it will break things.  You need some kind of translation
> layer if you're going to use NAND.
> 
> josh

I did not know that ftl.c is can be used also in 'Layer 2' instead
of dummy mtdblock.c. It's working. I confirmed that the ftl.c can 
be used on mtdram. I will test whether it works on a physical 
bare-nand flash memory. (maybe tomorrow)

Follow is what I have tested for a mtdram (64MB).

root@thermopylae:/mnt # modprobe mtdram

root@thermopylae:/mnt # modprobe mtdchar

root@thermopylae:/mnt # ftl_format /dev/mtd0
Partition size = 64 mb, erase unit size = 128 kb, 1 transfer units
Reserved 5%, formatted size = 62883328 bytes
Erasing all blocks...
+++++++*+++++++*+++++++*+++++++*+++++++*+++++++*+++++++*+++++++*
Writing erase unit headers...
format successful.

root@thermopylae:/mnt # modprobe ftl

root@thermopylae:/mnt # lsmod
Module                  Size  Used by
ftl                    11404  0
mtd_blkdevs            10368  1 ftl
mtdchar                 9092  0
mtdram                  3972  0
mtdcore                 7172  6 mtd_blkdevs,mtdchar,mtdram
autofs4                22788  1
3c59x                  43560  0
b44                    27276  0
ext3                  122504  1
jbd                    59048  1 ext3

root@thermopylae:/mnt # mkfs.ext3 /dev/ftla
mke2fs 1.38 (30-Jun-2005)
Filesystem label=
OS type: Linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
15360 inodes, 61408 blocks
3070 blocks (5.00%) reserved for the super user
First data block=1
Maximum filesystem blocks=62914560
8 block groups
8192 blocks per group, 8192 fragments per group
1920 inodes per group
Superblock backups stored on blocks:
        8193, 24577, 40961, 57345

Writing inode tables: done
Creating journal (4096 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 39 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.

root@thermopylae:/mnt # mount /dev/ftla /mnt/data0

root@thermopylae:/mnt # df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/hda1              29G  9.7G   18G  36% /
tmpfs                 507M     0  507M   0% /dev/shm
/dev/hdb1              95G  126M   90G   1% /mnt/data1
/dev/ftla              59M  5.2M   50M  10% /mnt/data0

-- 
Seongsu's personal blog - http://www.senux.com/
lp1 on fire (One of the more obfuscated kernel
messages)