Re: [PATCH 16/26] hw/display/apple-gfx: Asynchronous MMIO writes on x86-64

[BALATON Zoltan]

On Tue, 16 Jul 2024, Phil Dennis-Jordan wrote:
> On Mon, 15 Jul 2024 at 23:26, Philippe Mathieu-Daudé <philmd@linaro.org>
> wrote:
>
> > Hi Phil,
> >
> > On 15/7/24 23:06, Phil Dennis-Jordan wrote:
> > > This change ensures that the MMIO write calls into the PVG
> > > framework are performed asynchronously on a background dispatch
> > > queue. Without this, we rapidly run into re-entrant MMIO issues.
> > >
> > > This problem only seems to exist on x86-64 hosts. Conversely,
> > > doing it async on arm64/vmapple causes other issues,
> >
> > Such as?
>
> Sorry for being vague. I've just refreshed my memory by testing with async
> MMIO writes on aarch64, and the guest never manages to initialise the
> display at all. I've admittedly not attempted to debug through this in any
> significant way, though with PVG being something of a black box I'm not
> sure it's worth it. It works reliably on x86-64 with async writes, and on
> aarch64 with sync writes.

Only a guess but I think ARM like POWER has weak memory consistency so 
maybe some sync ops are needed between writes somewhere whereas it would 
work on X86_64 that has strong guarantees so no such explicit sync is 
needed? I may completely wrong though, it's just what this reminded me of.

Regards,
BALATON Zoltan