On 16 May 2006, at 11:12, Cliff Lawson wrote:
-----Original Message----- From: Jasmine Strong [mailto:jasmine@electronpusher.org]
Because 2.4 is obsolete and basing new projects on it is a very bad idea.
Methinks you missed the point. What I'm saying is PBL is happy to load a block of NAND out to a place in SDRAM at a location specified in that images module header and then jump to it's entry point if it happens to have the name "LINUX" in the module header.
I've actually written completely standalone LCD demo programs using this technique and called them "LINUX" and then the PBL has loaded/ executed them.
That image could be a self extracting 2.4 or 2.6 or, when available 2.8 Linux or perhaps even uBoot or anything your heart desires.
This is almost what Jonathan is doing - PBL loading littleLDR which loads uBoot. I seem to remember he tried the simpler form of getting PBL to load a chunk containing uboot directly (Jonathan, was that with a LINUX-named chunk or some other PBL loadable?) but ran into a size limit.
If we could have PBL load uboot directly then that would be preferable - Cliff, what does PBL expect from a LINUX block, any checksumminess or compression we have to worry about? (Although Jonathan's pre-uboot load shim seems to work now so it's probably moot anyway.)
Cheers,
Matt