On Tuesday 01 September 2009, Øyvind Harboe wrote: > Starting a new thread on a crazy idea I had to help alleviate > problems with coming up with the reset init sequence. > > Lots of hardware out there have working .elf files that set up > the hardware.
Called "boot loaders". And when OpenOCD is used to develop such boot loaders ... developers rely on scripted JTAG setup. I've seen such scripts top 200K, as produced by teams that do silicon verification. Trying all kinds of clock options, memory and bus timings, etc. Not in OpenOCD; but with other JTAG tools. > What if a known working elf image could be used instead of > having to first succeed in writing a reset init sequence? Makes me think of chickens and eggs debating which one crossed the street first. :) Also, that creating and maintaining libraries of such images is needless complication. But that it's clearly possible to do such things with OpenOCD today, if needed. In a way this is just what run_algorithm() does... > ... The trick with running an elf image > without a working reset init sequence is that the elf image > might be designed to run out of flash/rom. Using JTAG to download to ROM would be quite the trick! In many cases that code gets developed to run out of SRAM; e.g. loaded from a NAND or OneNAND bootblock as a second stage loader (mask ROM boot loader talks to NAND or a handful of other peripherals as the first stage). That level of simulation support is a lot of effort, and isn't very portable. Having a way to hook up to, for one option, QEMU might be a good addition to OpenOCD though. - Dave _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
