[RFC PATCH v2 2/2] ARM: signal: Fix unparseable iwmmxt_sigframe in uc_regspace[]

On Tue, Jun 27, 2017 at 11:08:12PM +0100, Russell King - ARM Linux wrote:

On Tue, Jun 27, 2017 at 06:04:07PM +0100, Dave Martin wrote:

quoted

In kernels with CONFIG_IWMMXT=y running on non-iWMMXt hardware, the
signal frame can be left partially uninitialised in such a way
that userspace cannot parse uc_regspace[] safely.  In particular,
this means that the VFP registers cannot be located reliably in the
signal frame when a multi_v7_defconfig kernel is run on the
majority of platforms.

The cause is that the uc_regspace[] is laid out statically based on
the kernel config, but the decision of whether to save/restore the
iWMMXt registers must be a runtime decision.

To minimise breakage of software that may assume a fixed layout,
this patch emits a dummy block of the same size as iwmmxt_sigframe,
for non-iWMMXt threads.  However, the magic and size of this block
are now filled in to help parsers skip over it.  A new DUMMY_MAGIC
is defined for this purpose.

It is probably legitimate (if non-portable) for userspace to
manufacture its own sigframe for sigreturn, and there is no obvious
reason why userspace should be required to insert a DUMMY_MAGIC
block when running on non-iWMMXt hardware, when omitting it has
worked just fine forever in other configurations.  So in this case,
sigreturn does not require this block to be present.

Reported-by: Edmund Grimley-Evans <redacted>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>

This looks fine to me.  Please drop it in the patch system, thanks.

Do you have a view on whether I should Cc-stable on this?

The patches seem to apply cleanly back to v3.4, but I'm not in a
position to test that far back easily.

As a reference point, Debian stretch seems to use v4.9.x for its
multiplatform distro kernel, so it may be worth going at least that
far back.  (jessie uses v3.16.x.)

Cheers
---Dave