Running Linux on a software model

Software models do not usually have all the features of a hardware platform like Ethernet, persistent Flash memory, boot monitor. This page describes a way to build an ELF image from kernel, filesystem and simple boot code to initialise the registers and UART0. The example files are to be used with the RealView^? models that emulate an Integrator/CP platform.

Kernel configuration requirements

The kernel needs the following configuration options set:

• CONFIG_BLK_DEV_RAM=y
• CONFIG_BLK_DEV_RAM_COUNT=16 (any number higher than 0)
• CONFIG_BLK_DEV_RAM_SIZE=16384 (should be at least the size of the filesystem image in KB)
• CONFIG_BLK_DEV_RAM_BLOCKSIZE=1024 (this is the default value)
• CONFIG_BLK_DEV_INITRD=y
• CONFIG_CMDLINE="root=/dev/ram0 initrd=0x00800000,10M ..." (initrd=... should specify the location and the size of the filesystem image)

Building the ELF model image

Copy the following files (attached) to a local directory:

• Makefile - builds the ELF model image
• boot.S (or boot-t2.S) - simple code to initialise some registers and UART0 before branching into the ARM (or Thumb-2) kernel
• model.lds - linker script used to generate the ELF model image

Copy the "uImage" kernel file and filesystem into the directory above and the "make" command will build a "linux-image.axf" file that can be loaded into the model and executed.

-- CatalinMarinas - 23 Jul 2007