Re: [linuxdc-dev] VMS flash handling

[Frederic S. <fre...@sx...>]

> So, in theory, you could hack something like RedBoot to be able to handle
> maple, and simply read in the kernel image one block at a time.. then map
> it into memory, and set the start address to the beginning of the mapped
> region, and boot from there.

Oh yes ... I said boot from VMU but code relocation (not mapping) into
memory before booting should be better.
 
> You wouldn't really get much benefit from this in my opinion, but oh well.

Flexibility.
 
> Writing is still a touchy subject. I've been working on an MTD driver for
> the VMU, though it interacts with the device over maple directly, and
> the block reads/writes are handled by an underlying flash mapping driver.

That is very interesting. But IMHO VMU flash capacity is very low.
Writing some code which is near a flash file system is a big job. I
think taht work around JFFS is a good example.
 
> The upside to this is that you can fully use it like any other embedded
> storage device, but you have to do it over maple.. which requires having
> the DC already booted and writing it from there.

This is definitely very interesting.
For me, needs are more about core kernel development and an easy way to
boot a custom one from flash (faster than read an image from serial
link).
 
> The only other option you have, is to try and dump to flash raw data that
> doesn't comply with the vmufs or any other filesystem, and then try and
> map that into memory.. that'd be the simplest way to go about it, but it
> renders the VMU pretty much useless for anything else until you can restore
> the vmufs.

It is absolutely that I want :)

Thanks for all these interesting information.

BRs,

Fred.