On Wed, Nov 24, 2010 at 10:24 AM, Scott Wood scottwood@freescale.com wrote:
On Wed, 24 Nov 2010 00:03:13 +0100 Wolfgang Denk wd@denx.de wrote:
Dear Scott Wood,
In message 20101123163204.4f843e61@udp111988uds.am.freescale.net you wrote:
How about playing with BATs before entering C code, so that the image always appears at the same effective address?
Not all systems have BATs.
I was speaking in the context of what he wanted to do with an 83xx board -- but the concept applies to any hardware with an MMU that isn't too painful to set up that early.
If someone wants to do this kind of thing on hardware that doesn't meet that description, that's another matter -- if the the hardware doesn't provide a nicer bank switching mechanism (e.g. p4080ds lets you rotate the flash banks' physical addresses, rather than change the reset vector), or an SRAM that U-Boot (or an SPL) could copy itself to before C code, etc.
It seems to me that we are applying at the architecture level a 'nice to have' which may belong at the board level. How many vendors are going to do a fancy 'two U-Boot images' trickery? Will it be (nearly) every 83xx board?
I agree with Scott - If you want to do something that fancy, provide support for it in your board hardware. I don't like the idea of diverging the core feature-set available at the architecture level if those features belong more at the SOC or Board level.
Regards,
Graeme