Re: [PATCH 4/8] edu: add smp_mb__after_rmw()

[Peter Maydell]

On Mon, 6 Mar 2023 at 15:06, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 3/6/23 15:24, Peter Maydell wrote:
> >>> Why is this device even messing around with multiple
> >>> threads and atomics anyway ??
> >> Because it is an example of deferring device work to another thread,
> >> just like on real hardware it may be deferred to an on-device
> >> microcontroller or CPU.
> > If we want to be able to do that, we should probably have
> > infrastructure and higher-level primitives for it that
> > don't require device authors to be super-familiar with
> > QEMU's memory model and barriers... The fact there are only
> > half a dozen other uses of qemu_thread_create() under hw/
> > suggests that in practice we don't really need to do this
> > very often, though.
>
> Yes, you're totally right about that.  I have never needed this kind of
> higher-level primitive so I haven't thought much about what it would
> look like.
>
> The usage of barriers isn't great, but all this patch does is correctness...

Yeah, I'm not objecting to this patch, to be clear. It's just
that it brought the edu device to my negative attention :-)

-- PMM