[ELDK] mmap: first byte always wrong

[Alberto Caballero]

Hi Stefano,

Yes, board flash is NOR-type. I will try to check the flash operation scheme
and also the use of the MTD partitions.

Regards,

Alberto

2010/11/26 Stefano Babic <sbabic at denx.de>

> On 11/26/2010 08:28 AM, Alberto Caballero wrote:
> > Hi,
>
> Hi,
>
> >
> > At my project, I will have a data configuration file stored in the board
> > Flash memory using an U-Boot filesystem image. At the initialization
> stage,
> > I read the image from the Flash and store it in the Linux filesystem.
> >
> > Of course at the U-Boot prompt I am able to check that image header at
> Flash
> > is correctly stored and all the bytes has the right values. I was
> wondering
> > if this has something to do with first access to mapped memory.
>
> You can try to read the flash at the required address twice and check if
> the first byte is always wrong. However, with mmap() you are bypassing
> the whole linux flash driver that it is responsible to provide the same
> interface independently from the underlying hardware.
>
> And again, I suppose you have NOR flash on board, because reading NAND
> flash with mmap simply does not work...
>
> >
> > When Linux initializes it creates a series of MTD partitions by its own.
>
> I do not think so. Linux creates partitions as you asked to do. Check
> your DTS file (I assume you are using a recent kernel) and modify the
> partition's size according to your requirement.
>
> > Now
> > the data in the Flash will be overlapped in two partitions so I will try
> to
> > use just one of them.
>
> Then it seems you have to adapt the configuration of the partitions.
>
> >
> > Regarding the mmap usage, I was able to check with another board with
> > another PowerPC processor using ELDK also, and in this case there is no
> > issue with Flash access.
>
> This does not depend on ELDK. Reading directly the flash chip can work
> on some hardware, but it could be unreliable with some other chips. For
> example, you do not know which was the last operation issued on the
> chip. If the flash driver has sent a CFI-Query, what you read is the
> internal mapping of the chip. Using the MTD subsystem you are sure that
> the flash driver will take care of all internal details.
>
> Best regards,
> Stefano Babic
>
> --
> =====================================================================
> DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
> =====================================================================
>