2 ms·

I wonder how long until the kernel is checking signatures on code pages as they're mapped into memory or modified (iOS style). Obviously no one wants the vendor

by dcow 9y ago

I wonder how long until the kernel is checking signatures on code pages as they're mapped into memory or modified (iOS style). Obviously no one wants the vendor gated platform lockdown aspects of any of this, but user-defined key whitelists for both kernel and userspace binaries would be really cool. I think I've heard Linus is not in favor of extending this notion to userspace binaries, or at least that he doesn't believe it should be the kernel's responsibility (for Linux anyway), though.

mjg59 9y ago
The kernel already has support for verifying signatures on binaries - check out IMA appraisal.
- dcow 9y ago
  Apropos: https://lwn.net/Articles/488906/ https://lwn.net/Articles/488906/ Hmm does this only apply to files though? I'm talking about kernel signature engorcement on any page mapped into memory.
  - mjg59 9y ago
    No, there's a presumption that any signed app isn't going to generate untrusted code
    - dcow 9y ago
      Thinking about it more that seems fair enough provided as a user I can operate the system in such a way that I'm only running trusted code.
jbit 9y ago
As well as the IMA Appraisal feature that mjg59 mentioned, you can also integrity check your entire filesystem with verity[1]. Android uses it to make sure the system partition isn't tampered with. [1] https://www.kernel.org/doc/Documentation/device-mapper/verity.txt https://www.kernel.org/doc/Documentation/device-mapper/verit...
- swiley 9y ago
  Because God forbid you tamper with the system partition. The security features of android are the best argument for the stance Linus takes on these things.