I sent this cleanup first... in mid-July (ugh). But better late than never, I guess.

From v1:
- fix buglets (thanks Gavin)
- add patch 3 on top

The kvm_{get,set}_one_reg functions have been around for a very long
time, and using them instead of open-coding the ioctl invocations
saves lines of code, and gives us a tracepoint as well. They cannot
be used by invocations of the ioctl not acting on a CPUState, but
that still leaves a lot of conversions in the target/arm code.

target/mips and target/ppc also have some potential for conversions,
but as I cannot test either (and they are both in 'Odd fixes' anyway),
I left them alone.

Survives some testing on a Mt. Snow.

Cornelia Huck (3):
  arm/kvm: convert to kvm_set_one_reg
  arm/kvm: convert to kvm_get_one_reg
  arm/kvm: convert to read_sys_reg64

 target/arm/kvm.c   |  28 +++-------
 target/arm/kvm64.c | 129 ++++++++++++---------------------------------
 2 files changed, 40 insertions(+), 117 deletions(-)

-- 
2.41.0

On Tue, 10 Oct 2023 at 15:25, Cornelia Huck <cohuck@redhat.com> wrote:
>
> I sent this cleanup first... in mid-July (ugh). But better late than never, I guess.
>
> From v1:
> - fix buglets (thanks Gavin)
> - add patch 3 on top
>
> The kvm_{get,set}_one_reg functions have been around for a very long
> time, and using them instead of open-coding the ioctl invocations
> saves lines of code, and gives us a tracepoint as well. They cannot
> be used by invocations of the ioctl not acting on a CPUState, but
> that still leaves a lot of conversions in the target/arm code.

I've applied patches 1 and 2 to target-arm.next.

thanks
-- PMM