6 ms·
There's a kernel-imposed boundary between being a user and being root. Right now there's also a partial barrier between being root and being the kernel that's f
by mjg59 9y ago
There's a kernel-imposed boundary between being a user and being root. Right now there's also a partial barrier between being root and being the kernel that's fairly easily circumvented. Now that we can protect the on-disk copy of the kernel from root, there's also interest in protecting the in-memory copy of the kernel from root.
- tjoff 9y agoYes. And lockdown helps with (right?): a) locking root out of kernel b) locking users from becoming root I interpret the reasoning as: if we can't do a AND b we won't bother with b. Which is confusing because b is many orders of magnitudes more important than a.
- mjg59 9y agob) is a barrier that's already supposed to exist, so these patches don't have an impact on it.
- tjoff 9y agoSupposedly yes, but shoudln't it vastly reduce the attack surface by not having to worry (compared to now) about all device drivers that use DMA (which otherwise is just impossible to keep track of). In my opinion reducing that attack surface alone on b) has a much higher impact than a) could ever achieve.
- Jasper_ 9y agoa) prevents a temporary breach of b) from turning into a persistent attack. b) is still very important, so important that when b) breaks it is considered a CVE!
- mjg59 9y agoThe stuff lockdown impacts is stuff that only root can do right now, so in theory it's only handling channels that are irrelevant to b). Users should already be prevented from triggering arbitrary DMA, for instance.
- tjoff 9y agoWell, should, again, device drivers are notoriously bad and some protection against that is extremely valuable.