[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S]()
From: |
Igor Mammedov |
Subject: |
Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S]() |
Date: |
Tue, 3 Oct 2017 17:01:39 +0200 |
On Tue, 3 Oct 2017 11:25:01 -0300
Eduardo Habkost <address@hidden> wrote:
> On Tue, Oct 03, 2017 at 02:14:46PM +0200, Igor Mammedov wrote:
> > it will help to remove code duplication in places
> > that currently open code registration of static
> > type[s] and remove necessity to declare function
> > for type_init() to call, when only static types
> > need to be registered.
> >
> > Signed-off-by: Igor Mammedov <address@hidden>
> > ---
> > I'm going to use it for CPU types in followup patches
> >
> > CC: address@hidden
> > ---
> > include/qemu/module.h | 10 ++++++++++
> > include/qom/object.h | 10 ++++++++++
> > qom/object.c | 9 +++++++++
> > 3 files changed, 29 insertions(+)
> >
> > diff --git a/include/qemu/module.h b/include/qemu/module.h
> > index 56dd218..60dd632 100644
> > --- a/include/qemu/module.h
> > +++ b/include/qemu/module.h
> > @@ -52,6 +52,16 @@ typedef enum {
> > #define type_init(function) module_init(function, MODULE_INIT_QOM)
> > #define trace_init(function) module_init(function, MODULE_INIT_TRACE)
> >
> > +#define REGISTER_STATIC_TYPES(t, nr)
> > \
>
> I'm unsure about the name. We already have a declarative way to
> register machine types (DEFINE_MACHINE), maybe use "DEFINE_*" for
> consistency?
>
> Also, this function won't work if t isn't a static variable, will
> it? The "STATIC" part looks redundant.
>
> In other words, how about "DEFINE_TYPE" and "DEFINE_TYPES"?
'define' word remind me more of declaration but here we actually
register types that's why 'register' looks better to me as it says
exactly what's going here.
>
>
> > +static void do_qemu_init_ ## t(void)
> > \
> > +{
> > \
> > + type_register_static_array(t, nr);
> > \
>
> qemu/module.h doesn't include qom/object.h. I think this should
> go to either qom/object.h or a new header file.
I could move it to qom/object.h
>
> > +}
> > \
> > +type_init(do_qemu_init_ ## t)
> > +
> > +#define REGISTER_STATIC_TYPE(t)
> > \
> > + REGISTER_STATIC_TYPES(t, 1)
> > +
> > #define block_module_load_one(lib) module_load_one("block-", lib)
> >
> > void register_module_init(void (*fn)(void), module_init_type type);
> > diff --git a/include/qom/object.h b/include/qom/object.h
> > index f3e5cff..17fcadd 100644
> > --- a/include/qom/object.h
> > +++ b/include/qom/object.h
> > @@ -789,6 +789,16 @@ Type type_register_static(const TypeInfo *info);
> > Type type_register(const TypeInfo *info);
> >
> > /**
> > + * type_register_static_array:
> > + * @infos: The array of the new type #TypeInfo structures.
> > + * @nr_infos: number of entries in @infos
> > + *
> > + * @infos and all of the strings it points to should exist for the life
> > time
> > + * that the type is registered.
> > + */
> > +void type_register_static_array(const TypeInfo *infos, int nr_infos);
>
> This new function is probably less controversial than the new
> module.h-based code. Maybe worth moving to a separate patch so
> it can be merged earlier?
>
> (If I were you, I would just use type_register_static_array()
> inside the existing *_cpu_register_types() functions by now, so
> this won't block the whole series, and worry about
> REGISTER_STATIC_TYPE*/DEFINE_TYPE* later.)
it would be nice to have both in one series so that
we won't have to touch/rewrite more than once code where
static cpu types are registered.
If you prefer I can do what you suggest and keep
*_cpu_register_types() functions
but I can't promise to return back to cleaning them up
or to REGISTER_STATIC_TYPE*/DEFINE_TYPE, so probably
someone else will have to pickup that work.
> > +
> > +/**
> > * object_class_dynamic_cast_assert:
> > * @klass: The #ObjectClass to attempt to cast.
> > * @typename: The QOM typename of the class to cast to.
> > diff --git a/qom/object.c b/qom/object.c
> > index 3e18537..40b1729 100644
> > --- a/qom/object.c
> > +++ b/qom/object.c
> > @@ -151,6 +151,15 @@ TypeImpl *type_register_static(const TypeInfo *info)
> > return type_register(info);
> > }
> >
> > +void type_register_static_array(const TypeInfo *infos, int nr_infos)
> > +{
> > + int i;
> > +
> > + for (i = 0; i < nr_infos; i++) {
> > + assert(type_register_static(&infos[i]));
>
> Why the assert()? type_register_static() is already guaranteed
> to never return NULL.
according to doc comment type_register_static might return 0 on failure
so qemu should die in that case.
>
> This will break if compiled with -DNDEBUG: assert() won't even
> evaluate the expression if NDEBUG is defined.
>
> > + }
> > +}
> > +
> > static TypeImpl *type_get_by_name(const char *name)
> > {
> > if (name == NULL) {
> > --
> > 2.7.4
> >
>
- [Qemu-devel] [PATCH 00/38] generalize parsing of cpu_model (part 2), Igor Mammedov, 2017/10/02
- [Qemu-devel] [PATCH 01/38] qom: add helper type_init_from_array(), Igor Mammedov, 2017/10/02
- Re: [Qemu-devel] [PATCH 01/38] qom: add helper type_init_from_array(), Philippe Mathieu-Daudé, 2017/10/02
- Re: [Qemu-devel] [PATCH 01/38] qom: add helper type_init_from_array(), Eduardo Habkost, 2017/10/02
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Eduardo Habkost, 2017/10/03
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](),
Igor Mammedov <=
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Eduardo Habkost, 2017/10/03
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Igor Mammedov, 2017/10/03
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Eduardo Habkost, 2017/10/03
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Peter Maydell, 2017/10/03
- Re: [Qemu-devel] [PATCH] qom: add helpers REGISTER_STATIC_TYPE[S](), Igor Mammedov, 2017/10/03
[Qemu-devel] [PATCH v2 0/3] add DEFINE_TYPES() helper to simplify batch type registration, Igor Mammedov, 2017/10/04
[Qemu-devel] [PATCH v2 1/3] qom: update doc comment for type_register[_static](), Igor Mammedov, 2017/10/04