Re: sysbus_create_simple Vs qdev_create

[Markus Armbruster <armbru@redhat.com>]

Paolo Bonzini <pbonzini@redhat.com> writes:

> On 29/07/20 15:18, Markus Armbruster wrote:
>>> Even code riddled by backwards-compatibility special cases, such as
>>> -accel and -machine, can share code between themselves and -object to
>>> some extent; this is thanks to functions such as object_property_parse,
>>> whose parsing is deferred to visitors and hence to QAPI.
>>
>> QOM relies on QAPI visitors to access properties.  There is no
>> integration with the QAPI schema.
>
> Indeed it doesn't use _all_ of the QAPI goodies.  It does use visitors
> and it's a major feature of QOM.

No argument.

>> Going through a visitor enables property access from QMP, HMP and CLI.
>> 
>> Access from C *also* goes through a visitor.  We typically go from C
>> type to QObject and back.  Comically inefficient (which hardly matters),
>> verbose to use and somewhat hard to understand (which does).
>
> It's verbose in the getters/setters, but we have wrappers such as
> object_property_set_str, object_property_set_bool etc. that do not make
> it too hard to understand.

qdev C layer:

    frob->prop = 42;

Least cognitive load.

QOM has no C layer.

qdev property layer works even when @frob has incomplete type:

    qdev_prop_set_int32(DEVICE(frob), "prop", 42);

This used to map property name to struct offset & copy the value.
Simple, stupid.

Nowadays, it is the same as

    object_property_set_int(OBJECT(frob), "frob", 42, &error_abort);

which first converts the int to a QObject, then uses a QObject input
visitor with a virtual walk to convert it back to int and store it in
@frob.  It's quite a sight in the debugger.

qdev "text" layer is really a QemuOpts layer (because that's what we had
back then).  If we have prop=42 in a QemuOpts, it calls

    set_property("prop", "42", frob, &err);

Nowadays, this is a thin wrapper around object_property_parse(),
basically

    object_property_parse(frob, "prop", 42, &err);

Fine print: except set_property() does nothing for @prop "driver" and
"bus", which look just like properties in -device / device-add, but
aren't.

object_property_parse() uses the string input visitor, which I loathe.

>> Compare to what QOM replaced: qdev.  Properties are a layer on top of
>> ordinary C.  From C, you can either use the C layer (struct members,
>> basically), or the property layer for C (functions taking C types, no
>> conversion to string and back under the hood), or the "text" layer
>> (parse from text / format to text).
>> 
>> My point is not that qdev was great and QOM is terrible.  There are
>> reasons we replaced qdev with QOM.  My point is QOM doesn't *have* to be
>> the way it is.  It is the way it is because we made it so.
>
> QOM didn't only replace qdev: it also removed the need to have a command
> line option du jour for any new concept, e.g. all the TLS stuff, RNG
> backends, RAM backends, etc.

Yes.  There are good reasons for QOM.

> It didn't succeed (at all) in deprecating chardev/netdev/device etc.,
> but this is a very underappreciated part of QOM, and this is why I think
> it's appropriate to say QOM is "C with classes and CLI/RPC
> serialization", as opposed for example to "C with classes and multi
> programming language interface" that is GObject.

That's fair.

>> I've long had the nagging feeling that if we had special-cased
>> containers, children and links, we could have made a QOM that was easier
>> to reason about, and much easier to integrate with a QAPI schema.
>
> That's at least plausible.  But I have a nagging feeling that it would
> only cover 99% of what we're doing with QOM. :)

The question is whether that 1% really should be done the way it is done
:)