Subject: Re: mapping shared memory at a fixed address
To: Rahul Kulkarni <firstname.lastname@example.org>
From: Bill Studenmund <email@example.com>
Date: 02/07/2005 13:39:15
Content-Type: text/plain; charset=us-ascii
On Mon, Feb 07, 2005 at 11:43:36AM -0800, Rahul Kulkarni wrote:
> I understand you could do relative pointers..
> But I think it still might be possible with some VM
> changes(hacks) to reserve a "virtual address region"
> in all address spaces, so as to prevent something else
> to map at that address in the first place, such that
> mmap always maps at the same requested address
> successfully? =20
You can't make things so you can be certain it won't happen. You can=20
however take steps to make failure vanishingly unlikely.
The easiest thing to try is to pick an address, like around 2 GB, that=20
won't be used in either process. Or maybe 1.5 GB. Then map there. Also,=20
map early in the process's life and chances are things will work.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)
-----END PGP SIGNATURE-----