> -----Original Message----- > From: Nathan Field [mailto:ndf@ghs.com] > Sent: Wednesday, February 04, 2004 4:36 PM > To: Shaun Savage > Cc: Prashant Viswanathan; 'linux-mips@linux-mips.org' > Subject: Re: loading kernel via EJTAG interface > > > > > I am trying to load a linux kernel through a EJTAG > device. I was told > > > that I should modify the kernel so that it doesnt attempt > to use the > > > parameters passed to it by the loader. However, I am not > quite sure as > > > to what it means and what exactly one has to do. I would really > > > appreciate any pointers/help/suggestions. > One approach is to create some sort of setup script that > "emulates" the boot loader. I've done that for a Malta board > (which uses a > boot loader called YAMON). Basically it does something like this: > > reset and run the board > sleep for 7 seconds to let YAMON do its normal setup > halt the board > do all the setup that YAMON would do when it runs a program > > That last step is where the magic happens. Basically I do things like > setup various registers to point to memory regions, and then in those > regions create arrays of pointers, which point to strings containing > things like the "environment" that YAMON fills in for programs and > arguments that it passes. Depending on the capabilities of > your debugging > environment this can either be easy or hard. My setup is really nice, > downloads for small kernels take about the same time as going through > tftp, but bigger kernels with a ramdisk are actually faster to push > through EJTAG, which is nice :) > > nathan Thanks for the input. I have a bootrom (different operating system) on my device and if I just boot up from the bootrom it would set up all the registers, controllers for me. Also, visionProbe (which I am using to download over the EJTAG interface) sets up all the controllers for me. So, can't I just load the kernel image and just start executing from "kernel_entry"? Prashant
