On Mon, 2020-05-04 at 13:22 +0200, Paolo Bonzini wrote:
> On 04/05/20 12:45, Stefan Hajnoczi wrote:
> > > @@ -983,7 +983,7 @@ static void device_set_realized(Object *obj, bool
> > > value, Error **errp)
> > > }
> > >
> > > assert(local_err == NULL);
> > > - dev->realized = value;
> > > + atomic_set(&dev->realized, value);
> >
> > A memory barrier is probably needed so that the atomic_read() thread
> > sees up-to-date dev fields.
>
> Yes, it should be a store-release for the false->true case. The
> true->false case probably doesn't matter as much.
On second thought, I think both cases matter, after I examined the device
removal case.
In device removal case, the device is first un-realized and then removed from
the bus,
so just like in device hotplug case, the scsi_device_find can give you an
unrealized device.
I will change this patch to set .realized to false at the start (if needed) of
the function and to true at the end (also if needed)
Will atomic_rcu_set work? or atomic_store_release?
(Both are the same thing, but former documents the purpose of using with RCU.
Best regards,
Maxim Levitsky
>
> Paolo
>
