> Its not _that_ difficult and its getting easier. Each platform
> that is a port knocks out more awkward spots. The mklinux/hppa
> port for example resolves stacks that grow up rather than down
> and the signal delivery with these.
Doesn't that eliminate most stack smashing exploits?
You can only hit function pointers, not the return address.
Most overwrites would fill empty stack space until the
stack limit is exceeded.
This looks better than preventing stack execution.
With the PowerPC binary compatibility, that would mean
the stack grows up on normal PowerMac Linux too, right?
I think I recall a flag in the Intel CPU to do that too.