Re: [RFC PATCH v3 02/11] plugins: add API for registering discontinuity callbacks

[Julian Ganz]

Hi Alex,

January 9, 2025 at 2:57 PM, "Alex Bennée" wrote:
> Julian Ganz <neither@nut.email> writes:
> >  diff --git a/include/qemu/qemu-plugin.h b/include/qemu/qemu-plugin.h
> >  index 9c67374b7e..f998a465e5 100644
> >  --- a/include/qemu/qemu-plugin.h
> >  +++ b/include/qemu/qemu-plugin.h
> >  @@ -273,6 +273,21 @@ QEMU_PLUGIN_API
> >  void qemu_plugin_register_vcpu_resume_cb(qemu_plugin_id_t id,
> >  qemu_plugin_vcpu_simple_cb_t cb);
> >  
> >  +/**
> >  + * qemu_plugin_register_vcpu_discon_cb() - register a discontinuity 
> > callback
> >  + * @id: plugin ID
> >  + * @cb: callback function
> >  + *
> >  + * The @cb function is called every time a vCPU receives a discontinuity 
> > event
> >  + * of the specified type(s), after the vCPU was prepared to handle the 
> > event.
> >  + * Preparation usually entails updating the PC to some interrupt handler 
> > or trap
> >  + * vector entry.
> > 
> The "usually" here is a bit of a weasel word. We should be clear what
> the contract is with the plugin. Can we say the PC will be updated to
> the next instruction that will execute after the callback?

The contract is indeed clear: the PC will always be updated to the
instruction that will be executed next, at least if we don't have a
second discontinuity (e.g. jump to an unmapped page). The "usually"
refers to the discontinuity itself: in the case of host calls, we don't
observe a "jump" and the next instruction executed will just be the
instruction following the call.

I could have phrased this better, and will make this more clear in the
next update.

Regards,
Julian