Re: [PATCH] arm64: signal: Update sigcontext reservations table

[Dave Martin <Dave.Martin@arm.com>]

On Wed, Jul 31, 2024 at 03:58:00PM +0100, Mark Brown wrote:
> On Wed, Jul 31, 2024 at 03:38:40PM +0100, Dave Martin wrote:
> 
> > For the patch itself, can you say whether the proposed update is right
> > or wrong, and whether you think we need to document this better and/or
> > change the approach?
> 
> I haven't double checked the sizeofs but it covers all the things in the
> frame and like I said before there's no exclusion between PSTATE.{SM,ZA}.
> I do think this should be clearer about what it's supposed to be
> tracking.
> 
> + * where vl is the non-streaming SVE vector length, as set with PR_SVE_SET_VL,
> + * and svl is the streaming SVE vector length, as set with PR_SME_SET_VL.
> 
> I'd just say VL is the vector length, that's the term the architecture
> uses and it says it's set with PR_SVE_SET_VL to clarify.

It's the worst-case sigframe size that we care about here, regardless
of what code a signal is delivered in the middle of.  Surely that
depends on both vector length settings?

PSTATE.SM and .ZA can be twiddled on and off in userspace by the
compiler IIUC; other translations units aren't supposed to care (or
notice), so we can't know ahead of time which vector length setting
will be used when generting the sigframe.  User code allocating the
stack must assume the worst.

Can you think of a description that clarifies this?

Cheers
---Dave