Re: GCC Compiler Induced Vulnerability - affects programs compiled with GCC 7 and 8 containing nested functions

[Jordan Glover <Golden_Miller83 () protonmail ch>]

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Tuesday, October 23, 2018 3:35 PM, Solar Designer <solar () openwall com> wrote:

> Here's a list of maybe-actionable items I came up with in response to
> Webroot's findings/reminder:
>
> 1.  More distros should start checking for executable stacks in program
>     binaries at package build time, and error out when this is unexpected.
>
> 2.  On Linux, we might want to have an enforcing mode (or several
>     sub-modes) in the kernel, where it'd keep the stack non-executable (and
>     possibly enforce W^X for other mappings as well), ignoring any flags in
>     the program binaries. I encouraged Vasiliy Kulikov to implement that
>     when he worked with us under GSoC 2011 on Linux kernel hardening tasks.
>     Here's the relevant thread, including a kernel patch:
>
>     https://www.openwall.com/lists/kernel-hardening/2011/07/18/8
>
>     For GCC trampolines to continue working, we can implement emulation of
>     the trampoline instructions like I introduced in -ow patches for 2.2.x
>     and like it's done in PaX/grsecurity. Vasiliy's patch includes that
>     (using code from PaX).
>
>     IIRC, we never actually submitted this upstream. Maybe the current
>     kernel hardening project (KSPP) should take and complete this effort.

There is S.A.R.A LSM[0] proposed by Salvatore Mesoraca with the aim to upstream
it to mainline kernel when needed infrastructure for it will be ready.

[0] https://sara.smeso.it/en/latest/

Jordan