Hi Sergey,

On Sun, Feb 17, 2008 at 10:15:57AM -0800, [EMAIL PROTECTED] wrote:
> > Thanks for pointing that patch out! I agree with Sergey that it
> > should be a run-time option. Perhaps some generic infrastructure
> > to support additional NAND commands could be leveraged upon
> > later.
> 
> The proper way to do this would be adding an additional command to NAND
> tools suite. Something like "nand write-ubl" or something.

Yep. If the write.jffs2 syntax is actually used and valid, it would
seem logical to have write.ubl too.

> > Personally, I think it's silly to have one tool to write one
> > part of flash and another tool to write the other, but I can
> > understand why people want to do it in different ways.
> 
> Those are totally different tools. Initial bootloader is used on virgin
> DaVinci hardware that doesn't have any software on it. Sure one can use
> U-Boot to write UBL and U-boot into NAND but one has to load that U-Boot
> somehow first.

Rudy's UBL flasher will boot u-boot over the UART. From there you
could download UBL & application code over the network.

Also, it only takes one unforgiving slip of the hand to cause
"nand erase\n" on unprotected flash to give you a virgin board,
so I claim it's not that much of a special case.

> And if an initial boot loader is required anyways (it is required
> to boot off of NAND anyways, U-Boot can NOT boot itself from NAND)
> why not use it for writing itself and user application (e.g.
> U-Boot) into NAND?

Because then you have two tools for writing different parts of flash
and a bloated UBL. Typically you'd have in flash the UBL (possibly
up to 5 times), an optional bootloader (possibly duplicated also), a
kernel, and a filesystem (perhaps two - one RO, one RW). There is
not yet a single tool that can do all of these, though there are
*many* tools that can do some of these.

> Another consideration is that U-Boot is not omnipresent and omnipotent. Yes,
> it is most frequently used user application but it is not mandatory. One can
> use something else (e.g. some standalone OS-less app, some OS kernel etc.)
> instead of U-Boot thus having no other options for writing into NAND...

I agree, however that shouldn't rule out the possibility for people
that do use U-boot.

> > From a quick Google and glance at the uboot code
> > (drivers/mtd/nand/nand_util.c), it seems there are already
> > differences in ECC formats between jffs2 and yaffs.
> 
> No, ECC is the same. ECC is on driver level and it is the same no matter
> what data is written to NAND. ECC takes just a _PART_ of OOB area and ECC
> code doesn't care what is written to unused OOB parts. That is where YAFFS
> and JFFS2 differ -- they use those _UNUSED_ OOB bytes differently.

Ahh, thanks for clarifying. Maybe write.ubl isn't the best option
from a code standpoint, but it probably still makes more sense to
the user.

My 2.2c.

Cheers,
Bernard.

_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source

Reply via email to