Re: [PATCH v2 7/7] arm64/sve: Don't zero non-FPSIMD register state on syscall by default

On Wed, Jul 20, 2022 at 10:20:23AM +0100, Will Deacon wrote:

On Tue, Jul 19, 2022 at 08:35:46PM +0100, Mark Brown wrote:

quoted

On Tue, Jul 19, 2022 at 06:35:37PM +0100, Catalin Marinas wrote:

quoted

quoted

quoted

The sysctl is disabled by default since it is anticipated that the risk
of disruption to userspace is low. As well as being within the
documented ABI this new behaviour mirrors the standard function call ABI
for SVE in the AAPCS which should mean that compiler generated code is
unlikely to rely on the current behaviour, the main risk is from hand
coded assembly which directly invokes syscalls. The new behaviour is
also what is currently implemented by qemu user mode emulation.

quoted

quoted

IIRC both Will and Mark R commented in the past that they'd like the
current de-facto ABI to become the official one. I'll let them comment.

quoted

That would be good.  I've not heard anything from Will either directly
or indirectly.  Mark R has indicated privately directly to me that he
originally pushed for the currently implemented behaviour and prefers
it.  Marc Zyngier has previously noted publicly the current behaviour
being a consideration in the context of discusion of optimisation ideas
like this one, I was a bit surprised that he commented on an earlier
patch in the series but not this one.  You have previously pushed back
on an attempt to document the current ABI, that was the main motivation
for writing this patch.

I remember discussing this somewhere at some point with somebody, but that's
not a tonne of use!

My opinion is that the observable behaviour is what matters. So if we
documented that the register state is "undefined" but it is in fact always
zero, then we should fix the documentation.

It's always zero with Linux, never zeroed by qemu user.

Does that help?

Yes, that's clear thanks though it does differ from what Catalin has
said about keeping the flexibility in the documentation - Catalin?