UBI bad block Management

UBI bad block Management

[Brijesh Singh]

Hi,
   The flash devices are getting of bigger in Size.The Init Scan time
will increase with bigger Size flash.How long are we going to rely on
MTD for Bad block Management.Even MTD doesn't maintain List but calls
the hardware, if hardware supports it.
This can be done in two ways,
1:) Straight right the bitmap,and if changed erase it.We can do it in
One reserved block.
           If allowed bad blocks are 20%,it will be erased 20% times
of total PEB count.
2:)Keep a Queue in Scan of Bad blocks.Ec header contains one field
called last bad block.Which is PEb no.So if block is found bad,next
block to be written first will have Bad block no in EC.So it is
maintained across reboots.
3:)Keep a Log in reserved Block.When it is getting full,create bitmap
and write it on the block.This will reduce the erasing of that block.

So we can maintain Bad block Map in UBI if it is necessary.

Re: UBI bad block Management

[Artem Bityutskiy]

Hi,

On Wed, 2007-02-14 at 14:58 +0530, Brijesh Singh wrote:
>    The flash devices are getting of bigger in Size.The Init Scan time
> will increase with bigger Size flash.How long are we going to rely on
> MTD for Bad block Management.

I hope forever. It is MTD's and only MTD's business to manage bad
blocks. Do not want MTD to do scanning? - Use on-flash BBT.

-- 
Best regards,
Artem Bityutskiy (Битюцкий Артём)

Re: UBI bad block Management

[Artem Bityutskiy]

Hi,

On Wed, 2007-02-14 at 14:58 +0530, Brijesh Singh wrote:
> Hi,
>    The flash devices are getting of bigger in Size.The Init Scan time
> will increase with bigger Size flash.How long are we going to rely on
> MTD for Bad block Management.Even MTD doesn't maintain List but calls
> the hardware, if hardware supports it.
> This can be done in two ways,
> 1:) Straight right the bitmap,and if changed erase it.We can do it in
> One reserved block.
>            If allowed bad blocks are 20%,it will be erased 20% times
> of total PEB count.
> 2:)Keep a Queue in Scan of Bad blocks.Ec header contains one field
> called last bad block.Which is PEb no.So if block is found bad,next
> block to be written first will have Bad block no in EC.So it is
> maintained across reboots.
> 3:)Keep a Log in reserved Block.When it is getting full,create bitmap
> and write it on the block.This will reduce the erasing of that block.
> 
> So we can maintain Bad block Map in UBI if it is necessary.

I really tried to realize what you mean, but failed, sorry. I do not
understand at all what are you about. May you re-phrase?

-- 
Best regards,
Artem Bityutskiy (Битюцкий Артём)

Re: UBI bad block Management

[Brijesh Singh]

Hi,
  I am satisfied by your last reply Artem.Seems Logical."It is MTD's
and only MTD's business to manage bad blocks."
  I had suggested three ways that would help UBI to maintain a Bitmap
of Bad Blocks.This would help speed up Initialization process.So if we
don't have on-flash BBT,maintaining a Bitmap will be an advantage.


On 2/14/07, Artem Bityutskiy <dedekind@infradead.org> wrote:
> Hi,
>
> On Wed, 2007-02-14 at 14:58 +0530, Brijesh Singh wrote:
> > Hi,
> >    The flash devices are getting of bigger in Size.The Init Scan time
> > will increase with bigger Size flash.How long are we going to rely on
> > MTD for Bad block Management.Even MTD doesn't maintain List but calls
> > the hardware, if hardware supports it.
> > This can be done in two ways,
> > 1:) Straight right the bitmap,and if changed erase it.We can do it in
> > One reserved block.
> >            If allowed bad blocks are 20%,it will be erased 20% times
> > of total PEB count.
> > 2:)Keep a Queue in Scan of Bad blocks.Ec header contains one field
> > called last bad block.Which is PEb no.So if block is found bad,next
> > block to be written first will have Bad block no in EC.So it is
> > maintained across reboots.
> > 3:)Keep a Log in reserved Block.When it is getting full,create bitmap
> > and write it on the block.This will reduce the erasing of that block.
> >
> > So we can maintain Bad block Map in UBI if it is necessary.
>
> I really tried to realize what you mean, but failed, sorry. I do not
> understand at all what are you about. May you re-phrase?
>
> --
> Best regards,
> Artem Bityutskiy (Битюцкий Артём)
>
>