On Thu, Sep 02, 2021 at 08:54:26AM -0700, Keith Packard wrote:
> Placing thread_info in the kernel stack leaves it vulnerable to stack
> overflow attacks. This short series addresses that by using the
> existing THREAD_INFO_IN_TASK infrastructure.
Very cool! Thanks for working on this. If you want, you can refer to the
KSPP bug for this too:
https://github.com/KSPP/linux/issues/1
(Anyone want to do MIPS?)
-Kees
--
Kees Cook
On Thu, Sep 02, 2021 at 09:07:42AM -0700, Kees Cook wrote:
> On Thu, Sep 02, 2021 at 08:54:26AM -0700, Keith Packard wrote:
> > Placing thread_info in the kernel stack leaves it vulnerable to stack
> > overflow attacks. This short series addresses that by using the
> > existing THREAD_INFO_IN_TASK infrastructure.
>
> Very cool! Thanks for working on this. If you want, you can refer to the
> KSPP bug for this too:
> https://github.com/KSPP/linux/issues/1
Not so fast. It's buggy. I've rejected this "solution" before.
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
On Thu, 2 Sept 2021 at 18:07, Kees Cook <[email protected]> wrote:
>
> On Thu, Sep 02, 2021 at 08:54:26AM -0700, Keith Packard wrote:
> > Placing thread_info in the kernel stack leaves it vulnerable to stack
> > overflow attacks. This short series addresses that by using the
> > existing THREAD_INFO_IN_TASK infrastructure.
>
> Very cool! Thanks for working on this. If you want, you can refer to the
> KSPP bug for this too:
> https://github.com/KSPP/linux/issues/1
>
> (Anyone want to do MIPS?)
>
I take it this breaks the GCC plugin based per-task stack protector,
given that it emits code to mask the stack pointer and apply an offset
to the resulting value.
It would be nice if we could replace this with something suitable for
THREAD_INFO_IN_TASK, and if it is suitable enough, try and get the
GCC/Clang folks to adopt it as well (which was never going to happen
for the stack pointer mask/offset approach)
Where can I find these patches? I don't see them on linux-arm-kernel@
On Thu, Sep 02, 2021 at 06:18:29PM +0200, Ard Biesheuvel wrote:
> On Thu, 2 Sept 2021 at 18:07, Kees Cook <[email protected]> wrote:
> >
> > On Thu, Sep 02, 2021 at 08:54:26AM -0700, Keith Packard wrote:
> > > Placing thread_info in the kernel stack leaves it vulnerable to stack
> > > overflow attacks. This short series addresses that by using the
> > > existing THREAD_INFO_IN_TASK infrastructure.
> >
> > Very cool! Thanks for working on this. If you want, you can refer to the
> > KSPP bug for this too:
> > https://github.com/KSPP/linux/issues/1
> >
> > (Anyone want to do MIPS?)
> >
>
> I take it this breaks the GCC plugin based per-task stack protector,
> given that it emits code to mask the stack pointer and apply an offset
> to the resulting value.
>
> It would be nice if we could replace this with something suitable for
> THREAD_INFO_IN_TASK, and if it is suitable enough, try and get the
> GCC/Clang folks to adopt it as well (which was never going to happen
> for the stack pointer mask/offset approach)
I'd love to see the native GCC offset stuff work on arm32, but it's not
clear to me how much work that would be. It's implemented for several
architectures already. I've tried to capture the matrix here:
https://github.com/KSPP/linux/issues/29
-Kees
--
Kees Cook