Le Thu, Nov 22, 2001 at 08:35:44PM -0500, Sunil Beta a �crit :
> The loader is written and locked into the flash. The kernel is written
> without locking afaik and the same goes for the ramdisk. IMO you could try
> to force writing the Ramdisk, where you should get the same kinda error.
Yes, I get.

> Then the obvious change you need is to patch Hermit to write to the flash
> on that board. It's got Intel L28F320 Flash devices in Rev B. Maybe
L28F320, yes, it is the same on my board.

> there's a difference in Rev A and hence the checksumming trouble. I
> remember having a problem with Hermit when I was downloading my kernel,
> but that was because of an improperly implemented function to display
> addresses and sizes of data being written in the current block. 
-> This makes me have a look at memory map:

In the doc provided with the board, I read this:
"Keep in mind that CS89712 on-board flash memory has 78 non-uniform sized
sectors. Sectors 0-7 and 39-46 are each 16 KB, while sectors 8-38 and
47-77 are 128 KB."

8 16k-blocks, 31 128k-blocks, 8 16k-blocks, 31 128k-blocks = 8Mo

'hermit  --anonymous-regions  map' lets me see 8 16k blocks
followed by 63 128k blocks.
8 16k-blocks, 63 128k-blocks = 8Mo

src/target/cs89712/memmap.c
static const char flash_blocking[] = "2x16K/l,6x16K,63x128K";

I'll look in that direction...

thanks
seb

_______________________________________________
http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
Please visit the above address for information on this list.

Reply via email to