Rene Mayrhofer wrote: > I will be using squashfs + aufs for our upcoming version 3.0 of > Gibraltar firewall. For security reasons, aufs is not used as root > filesystem (the squashfs is mounted as root), but multiple aufs mounts > are used for partial overlays where write access is necessary (/etc and > /var by default and sub-tree of /lib, /usr, /bin, and /sbin when patches > are applied at run-time). This makes it harder for potential (scripted) > attackers as most parts of the root tree are read-only. >
Have you considered an approach which DOES use aufs for the rootfs, but stacks only read-only images? The idea is to break the system up into some read-only modules which can be stacked together. The top level always needs to be slightly writable, but of course I accept this can be solved in lots of ways including symlinks to a writable partition, aufs, etc > In terms of run-time patching, the current script creates tmpfs/aufs > specifically for those directories where changes are being made (and the > patch packages are gpg-signed). I would be interested in how/where you check the signing - this is something I wanted to add to my setup, but was worried about the boot time slowdown? Is it straightforward to get gpg going in the initrd..? > However, I have also been thinking of a > patch/package format that could be "loaded" more efficiently during > bootup. We might want to defined a pseudo-standard here... (I already > have a fairly extensible tgz/gpg based format that might be sufficient > for patches and addon-type packages). > Interested to hear more about where you are going with that idea? What are you doing right now? Thanks for sharing Ed W ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july
