Re: [Qemu-devel] [PATCH] hw/s390x/sclp: Mark the sclp device with user_creatable = false

[Cornelia Huck]

On Wed, 4 Oct 2017 19:06:57 +0200
Thomas Huth <address@hidden> wrote:

> On 04.10.2017 17:18, Pierre Morel wrote:
> > On 04/10/2017 15:53, Thomas Huth wrote:  
> >> The "sclp" device is just an internal device that can not be instantiated
> >> by the users. If they try to use it, they only get a simple error
> >> message:
> >>
> >> $ qemu-system-s390x -nographic -device sclp
> >> qemu-system-s390x: Option '-device s390-sclp-event-facility' cannot be
> >> handled by this machine
> >>
> >> Since sclp_init() tries to create a TYPE_SCLP_EVENT_FACILITY which is
> >> a non-pluggable sysbus device, there is really no way that the "sclp"
> >> device can be used by the user, so let's set the user_creatable = false
> >> accordingly.
> >>
> >> Signed-off-by: Thomas Huth <address@hidden>
> >> ---
> >>   hw/s390x/sclp.c | 5 +++++
> >>   1 file changed, 5 insertions(+)
> >>
> >> diff --git a/hw/s390x/sclp.c b/hw/s390x/sclp.c
> >> index 30aefbf..9be0cb8 100644
> >> --- a/hw/s390x/sclp.c
> >> +++ b/hw/s390x/sclp.c
> >> @@ -606,6 +606,11 @@ static void sclp_class_init(ObjectClass *oc, void
> >> *data)
> >>       dc->realize = sclp_realize;
> >>       dc->hotpluggable = false;
> >>       set_bit(DEVICE_CATEGORY_MISC, dc->categories);
> >> +    /*
> >> +     * Reason: Creates TYPE_SCLP_EVENT_FACILITY in sclp_init
> >> +     * which is a non-pluggable sysbus device
> >> +     */
> >> +    dc->user_creatable = false;
> >>
> >>       sc->read_SCP_info = read_SCP_info;
> >>       sc->read_storage_element0_info = read_storage_element0_info;
> >>  
> > 
> > I must miss something.
> > Why is the sclp device not a SYS_BUS_DEVICE ?
> > The problem seems to come from the heterogeneity of the sclp and sclp
> > generated devices.  
> 
> I wonder whether it should rather be the other way round: Why is
> TYPE_SCLP_EVENT_FACILITY a sysbus device? Sysbus is an abstraction for
> simple on-board devices, while TYPE_SCLP_EVENT_FACILITY seems rather to
> be a pseudo-device... so it should rather simply be of TYPE_DEVICE
> instead? Or do I miss something?

We do have a number of devices that are sysbus because you could not
have bus-less devices back then. We might want to convert them to a
simple TYPE_DEVICE, but this needs double-checking regarding resets and
other things and I'd rather not spend that much time on that right now.