3 ms·
All this seems like a pretty compelling reason to move to using VMs and type safety to provide process isolation instead of using the hardware, e.g. https://www
by jordanthoms 9y ago
All this seems like a pretty compelling reason to move to using VMs and type safety to provide process isolation instead of using the hardware, e.g. https://www.destroyallsoftware.com/talks/the-birth-and-death-of-javascript https://www.destroyallsoftware.com/talks/the-birth-and-death... or Microsoft's Singularity and Midori projects.
It's a shame there's so much inertia behind the current setup of hardware memory management etc that it seems it'll be a long time before anything actually happens here.
- runeks 9y agoHow would this approach have helped in this case? As I understand it, the flaw in question allows reading kernel memory by executing user space code. How can a layer of software, on top of this type of buggy hardware, fix this issue? In one sense, the issue has already been fixed by a layer of software on top — which restricts a bunch of stuff, and reduces performance — but I assume this isn’t what you’re looking for.
- jordanthoms 9y agoThose systems don't use the buggy aspect of the hardware (memory protection) at all. Instead, all code is run inside a VM which provides memory protection and process isolation - there is no 'native' code at all. Not using the hardware memory protection provides a ~20% performance boost, which makes up for the ~20% overhead of running everything through VMs.