norman.feske at ...1...
Tue Aug 4 13:08:05 CEST 2009
Sven Fülster wrote:
> is there a special reason why you don't have implemented the
> 'attach_at'-support to a local address in the base-linux code?
> I need this feature (and my fix seems to work well here).
in constrast to the L4 platforms, on which 'attach_at' is needed
for constructing child address spaces, we had no use for 'attach_at'
on Linux so far. We simply start processes using 'execve' and leave
the allocation of virtual memory to the Linux kernel.
> I'll also need this feature in the base-okl4 code (and I would try to
> add it also there if necessary...)
It is there already. ;-)
> Is there any risk/ nasty surprise when using it?
On Linux, you'll have to make assumptions about how the Linux
kernel allocates mmap regions. I would try to avoid relying on such
heuristics. Do you have a particular reason for specifying local
addresses for 'mmap()'?
More information about the users