Re: [PATCH 19/26] rust: rename qemu-api modules to follow C code a bit more

[Paolo Bonzini]

On 12/12/24 10:52, Zhao Liu wrote:
> On Mon, Dec 09, 2024 at 01:37:10PM +0100, Paolo Bonzini wrote:
> > Date: Mon,  9 Dec 2024 13:37:10 +0100
> > From: Paolo Bonzini <pbonzini@redhat.com>
> > Subject: [PATCH 19/26] rust: rename qemu-api modules to follow C code a bit
> >   more
> > X-Mailer: git-send-email 2.47.1
> >
> > A full match would mean calling them qom::object and hw::core::qdev.  For now,
> > keep the names shorter but still a bit easier to find.
> >
> > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> > ---
> >   rust/hw/char/pl011/src/device.rs              |  4 +-
> >   rust/qemu-api-macros/src/lib.rs               |  2 +-
> >   rust/qemu-api/meson.build                     |  5 +-
> >   rust/qemu-api/src/lib.rs                      |  5 +-
> >   rust/qemu-api/src/module.rs                   | 43 +++++++++++
> >   rust/qemu-api/src/prelude.rs                  |  2 +-
> >   .../qemu-api/src/{device_class.rs => qdev.rs} |  4 +-
> >   rust/qemu-api/src/{definitions.rs => qom.rs}  | 74 +++++++++----------
> >   rust/qemu-api/src/sysbus.rs                   |  2 +-
> >   rust/qemu-api/tests/tests.rs                  |  5 +-
> >   10 files changed, 92 insertions(+), 54 deletions(-)
> >   create mode 100644 rust/qemu-api/src/module.rs
> >   rename rust/qemu-api/src/{device_class.rs => qdev.rs} (97%)
> >   rename rust/qemu-api/src/{definitions.rs => qom.rs} (83%)
> >
> > --- a/rust/qemu-api/src/definitions.rs
> > +++ b/rust/qemu-api/src/qom.rs
> > @@ -2,7 +2,37 @@
> >   // Author(s): Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
> >   // SPDX-License-Identifier: GPL-2.0-or-later
> >   
> > -//! Definitions required by QEMU when registering a device.
> > +//! Bindings to access QOM functionality from Rust.
> > +//!
> > +//! This module provides automatic creation and registration of `TypeInfo`
> > +//! for classes that are written in Rust, and mapping between Rust traits
> > +//! and QOM vtables.
> > +//!
> > +//! # Structure of a class
> > +//!
> > +//! A concrete class only needs a struct holding instance state. The struct 
> > must
> > +//! implement the [`ObjectType`] and [`IsA`] traits, as well as any `*Impl`
> > +//! traits provided by its superclasses.
>
> In this commit, this comment is a bit ahead, but I think it's okay.
>
> qom and qdev are both good names. In addition, we can rename the files
> of PL011 as well. Perhaps device_class.rs could be merged into device.rs
> (and eventually renamed to pl011.rs). I guess you might be planning to
> keep it until the cleanup of vmstate and property is done.

Yeah, I don't have any specific plans but memory_ops will certainly go 
away.  device_class doesn't do much, but keeping it separate is a 
reminder for things that are still there to be cleaned up.

As to VMState, there are two parts.  One is the vmstate_description 
macro, probably it has to be replaced with something else to incorporate 
the trampolines for pre_save/post_load/...  I haven't looked at it but 
it should not be a lot of work.

The second is VMStateFields, for which my idea is to implement a trait 
on types to retrieve a basic VMStateField (for example something like 
vmstate_uint32 would become an implementation of the VMState trait on 
u32).  Then you'd write something like "vmstate_of!(Type, 
field).with_version_id(2)" (i.e. vmstate_of retrieves the basic field 
and fills in the offset, then you apply more changes on top).  But that 
may take a while, and I think it cannot be done without the 
const_refs_to_static feature, which is only stable in 1.83.0.

Paolo