[U-Boot] Need Help in bulding U-boot-2010-06 for OpenRD client

[Albert ARIBAUD]

Le 08/07/2010 08:09, Prafulla Wadaskar a écrit :
>
>
>> -----Original Message-----
>> From: Albert ARIBAUD [mailto:albert.aribaud at free.fr]
>> Sent: Wednesday, July 07, 2010 7:36 PM
>> To: Prafulla Wadaskar
>> Cc: u-boot at lists.denx.de; kalyan karnati
>> Subject: Re: [U-Boot] Need Help in bulding U-boot-2010-06 for
>> OpenRD client
>>
>> Le 07/07/2010 13:44, Prafulla Wadaskar a écrit :
>>
>>> you should use u-boot.bin to load in RAM @ 0x600000 (TEXT_BASE)
>>
>> That's what I do, either through JTAG or by tftp/go in
>> resident u-boot.
>>
>>> for boot from RAM, you must need at least SDRAM configured,
>> so JTAG is the way
>>
>> JTAG does work indeed. Howvever, when:
>>
>> - I power up the board;
>> - ROM boot loads u-boot from NAND and starts it;
>> - I stop u-boot and get to the prompt;
>> - I 'tftp' an u-boot.bin image into RAM at 1000000;
>> - I 'go' to 1000000
>>
>> ... the tftp'ed image fails to boot although SDRAM is
>> obviously already
>> configured.
>
> This is obvious, since u-boot being copied is not created with TEXT BASE=0x100000, is it?
> And I don’t know how "go TEXT BASE" responds to u-boot binary itself, I have never tested this use case.

Well, it isn't so obvious, as normally one should be able to start 
u-boot from any address, as the first thing u-boot does is relocate 
itself back to its TEXT_BASE (see arch/arm/cpu/arm926ejs/start.S). This 
feature is precisely there to allow chaining one u-boot in RAM from 
another one. This relocation feature is compiled in by default; one must 
define CONFIG_SKIP_RELOCATE_UBOOT, to explicitely forbid u-boot from 
doing so.

>> OTOH, openOCD boots the RAM image fine, so I guess OpenOCD does some
>> inits that the boot rom plus resident u-boot don't.
>
> Openocd configures RAM before loading u-boot.bin
> U-boot.kwb does the same but internal bootROM does this instead of openOCD in case of boot from NAND
> U-boot.bin anyway needs SDRAM configurations to be done by someone else before starting execution.

I don't mean SDRAM is not initialized; obviously it is. However, 
something else prevents the kirkwood u-boot to chain from itself. I'd 
thought an initialization is missing, but actually it could be something 
about the relocation.

> Regards..
> Prafulla . .

Amicalement,
-- 
Albert.