Re: [PATCH 0/2] ARM: prctl: Reject PR_SET_MDWE where not supported

From: Zev Weiss
Date: Thu Feb 08 2024 - 02:42:53 EST

Hi Helge,

Thanks for taking a look!

On Wed, Feb 07, 2024 at 11:02:24PM PST, Helge Deller wrote:
Hi Zev,

On 2/8/24 02:26, Zev Weiss wrote:

I noticed after a recent kernel update that my ARM926 system started
segfaulting on any execve() after calling prctl(PR_SET_MDWE). After
some investigation it appears that ARMv5 is incapable of providing the
appropriate protections for MDWE, since any readable memory is also
implicitly executable.

(Note that I'm not an expert in either ARM arch details or the mm
subsystem, so please bear with me if I've botched something in the
above analysis.)

The prctl_set_mdwe() function already had some special-case logic
added disabling it on PARISC (commit 793838138c15, "prctl: Disable
prctl(PR_SET_MDWE) on parisc"); this patch series (1) generalizes that
check to use an arch_*() function, and (2) adds a corresponding
override for ARM to disable MDWE on pre-ARMv6 CPUs.

Instead of splitting it out to a new function in mman.h,
I'd prefer having it as config option, e.g. ARCH_HAS_NO_MDWE_SUPPORT (?)
which could be checked instead.
For parisc we still want to allow mdwe in the future, we just have
to wait until most user-space programs have updated to the latest
binaries which don't need an executable stack any longer.

I considered that, but it seems that ARM kernels at least may not know the answer to that question at compile-time -- see patch 2, where the ARM implementation does a runtime check on cpu_architecture().

With the series applied, prctl(PR_SET_MDWE) is rejected on ARMv5 and
subsequent execve() calls (as well as mmap(PROT_READ|PROT_WRITE)) can
succeed instead of unconditionally failing; on ARMv6 the prctl works
as it did previously.

Since this was effectively a userspace-breaking change in v6.3 (with
newer MDWE-aware userspace on older pre-MDWE kernels the prctl would
simply fail safely) I've CCed -stable for v6.3+, though since the
patches depend on the PARISC one above it will only apply cleanly on
the linux-6.6.y and linux-6.7.y branches, since at least at time of
writing the 6.3 through 6.5 branches don't have that patch backported
(due to further missing dependencies [0]).

I think you don't need to worry about that, since stable kernel series
for 6.3 up to 6.5 were stopped...

Ah, hadn't realized that -- thanks for the tip.