Fixed memory address

[Bernardo Innocenti]

On 10/05/2007 03:28 AM, Bert Freudenberg wrote:

> Would that be possible with the virtualization containers?

I don't think vservers messes around with the virtual
address space of processes.


> Btw, if I'm not mistaken, the Linux loader does something similar, so  
> if the relocating step could be skipped, it may help every executable  
> in the system.

Exactly.  This is what prelink does on F7.

I think we could run prelink on the OLPC images from
within pilgrim, just before converting it to jffs2.

It will result in a slight speedup in startup time, but
nothing compared to importing modules that dominates
startup time of the Python activities.


> Or is that already possible because of virtual memory, and the base  
> addresses are just randomized for security reasons? We wouldn't need  
> that kind of security with rainbow, do we?

The base address of processes are randomized for
security on several Linux distributions, including
Fedora.

You can disable it globally by doing this:

 echo 0 >/proc/sys/kernel/randomize_va_space


But this, of course, it not an option on the OLPC.  You can
also disable it by setting a flag in your ELF binary.
There's a tool to do that, but I can't remember its name
right now.  It was documented somewhere in the RedHat
release notes when they introduced va space randomization.

-- 
 \___/
 |___|  Bernardo Innocenti - http://www.codewiz.org/
  \___\ One Laptop Per Child - http://www.laptop.org/