mtd_info/mtd_info_user>size wrong type

mtd_info/mtd_info_user>size wrong type

[Charles Manning]

Folks
As people are starting to use larger NAND partitions (4GB+), an inconsistency 
in the size field has become apparent.

This is a unint32_t which is clearly too small for modern NAND arrays.

The read/write funcions use loff_t which is correct.

It would seem that size needs to be increated to loff_t too.

-- Charles

Re: mtd_info/mtd_info_user>size wrong type

[Josh Boyer]

On 4/6/07, Charles Manning <manningc2@actrix.gen.nz> wrote:
> Folks
> As people are starting to use larger NAND partitions (4GB+), an inconsistency
> in the size field has become apparent.
>
> This is a unint32_t which is clearly too small for modern NAND arrays.

Yes, known issue.

> The read/write funcions use loff_t which is correct.
>
> It would seem that size needs to be increated to loff_t too.

Have you tried doing that?  I seem to recall hearing that it was a bit
more involved than just changing the size member to be 64 bit.

josh

Re: mtd_info/mtd_info_user>size wrong type

[Charles Manning]

On Saturday 07 April 2007 13:26, Josh Boyer wrote:
> On 4/6/07, Charles Manning <manningc2@actrix.gen.nz> wrote:
> > Folks
> > As people are starting to use larger NAND partitions (4GB+), an
> > inconsistency in the size field has become apparent.
> >
> > This is a unint32_t which is clearly too small for modern NAND arrays.
>
> Yes, known issue.
>
> > The read/write funcions use loff_t which is correct.
> >
> > It would seem that size needs to be increated to loff_t too.
>
> Have you tried doing that?  I seem to recall hearing that it was a bit
> more involved than just changing the size member to be 64 bit.

I have not tried this though someone is aty present. I do know that someone 
has used YAFFS with Linux on a 6G or so partition.However they wanted a 
really fast nand driver and wrote a custom nand driver.

There definitely are some downstream problems too. It is not just a matter of 
changing the type.

eg using do_div() instead of divides and making sure that the right values 
flow through the whole  mtd subsystem fron nand_base..c right through 
mtdpart, mtdcat, the tools etc.

-- Charles