Post by Washington RatsoPost by UnruhPost by Washington RatsoI am running Linux 2.6.26 on a PPC. I want to boot linux without using
a bootloader to speed up the boot process but, due to the size of the
kernel, I cannot put it at the reset vector. I was told that I have
to put a branch instruction at the reset vector location and branch to
the location where my kernel is.
No idea what you are talking about. Whatever you do, you need a bootloader.
The kernel must be put into memory. Now if you do not like the 10 sec
delay, put in a 1 sec delay, or 0 sec delay. (Not advised since that means
you can never get at the boot sequence and alter things if something goes
wrong).
Post by Washington RatsoHow do I put a branch instruction at the reset vector?
Why in the world would you want to write your own bootloader? You do not
have enough to do in your life, but want to spend half a year doing
something others have already done far better?
I was told bu a consultant that it is possible to boot Linux in about
3 seconds. All one has to do is put vmlinux.gz at the reset vector.
He said he has done this. No bootloader is needed. But, I cannot put
vmlinux.gz at the reset vector because there is not enough space for
me to put vmlinux.gz there. So, he said to put a branch instruction
at the reset vector that would say to branch to an address where I
could put vmlinux.gz.
Since you are posting this in comp.arch.embedded, I'll assume the
consultant meant to put the images into ROM and boot from ROM. Even
decompressing and loading to RAM will be a lot faster reading ROM than
reading the image from disc. You would still need a boot loader.
As far as setting the boot vector, try reading the documentation of
your motherboard hardware. Changing the boot vector possibly means
changing the BIOS, or changing basic hardware settings. I'm guessing
changing the BIOS (ROM).
Ed