[PATCH v1] KVM: arm64: Allocate for maximum vector length for pKVM host sve state

[Mark Brown]

On Wed, Jun 05, 2024 at 04:03:12PM +0100, Fuad Tabba wrote:

> +	/* The maximum vl encountered for any cpu in the system. */
> +	int max_system_vl;
> +

As previously mentioned I'd really prefer not to call this _system_ due
to the potential confusion with the maximum VL we're using.  Naming is
hard.

> @@ -1031,7 +1031,12 @@ static void vec_probe_vqs(struct vl_info *info,
>  
>  		vq = sve_vq_from_vl(vl); /* skip intervening lengths */
>  		set_bit(__vq_to_bit(vq), map);
> +
> +		if (!max_vl)
> +			max_vl = vl;
>  	}
> +
> +	info->max_system_vl = max((int) max_vl, info->max_system_vl);

It's not clear to me why we stage this in a local variable prior to
updating the global value, and like I say it does feel cleaner and safer
to keep the current split between enumeration of the PE and integration
of the results.  This should *work* but it doesn't feel consistent with
the rest of the code.

> -               kvm_host_sve_max_vl = sve_max_vl();
> +               kvm_host_sve_max_vl = sve_max_system_vl();

I appreciate that we're adding a host_ on the front of it but it would
feel safer to keep the name of the hypervisor copy of the value
consistent with where we copied it from, sve_max_vl is a different thing
in the host kernel.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20240605/089b8563/attachment.sig>