Re: [PATCH] migration: Move populate_vfio_info() into a separate file

[Juan Quintela <quintela@redhat.com>]

"Dr. David Alan Gilbert" <dgilbert@redhat.com> wrote:
> * Thomas Huth (thuth@redhat.com) wrote:
>> On 15/03/2021 22.05, Philippe Mathieu-Daudé wrote:
>> > Hi Thomas,
>> > 
>> > +Alex
>> > 
>> > On 3/15/21 8:07 PM, Thomas Huth wrote:
>> > > The CONFIG_VFIO switch only works in target specific code. Since
>> > > migration/migration.c is common code, the #ifdef does not have
>> > > the intended behavior here. Move the related code to a separate
>> > > file now which gets compiled via specific_ss instead.
>> > > 
>> > > Fixes: 3710586caa ("qapi: Add VFIO devices migration stats in
>> > > Migration stats")
>> > > Signed-off-by: Thomas Huth <thuth@redhat.com>
>> > > ---
>> > >   migration/meson.build |  3 ++-
>> > >   migration/migration.c | 15 ---------------
>> > >   migration/migration.h |  2 ++
>> > >   migration/special.c   | 25 +++++++++++++++++++++++++
>> > >   4 files changed, 29 insertions(+), 16 deletions(-)
>> > >   create mode 100644 migration/special.c
>> > > 
>> > > diff --git a/migration/meson.build b/migration/meson.build
>> > > index 9645f44005..e1f72f6ba0 100644
>> > > --- a/migration/meson.build
>> > > +++ b/migration/meson.build
>> > > @@ -30,4 +30,5 @@ softmmu_ss.add(when: ['CONFIG_RDMA', rdma],
>> > > if_true: files('rdma.c'))
>> > >   softmmu_ss.add(when: 'CONFIG_LIVE_BLOCK_MIGRATION', if_true:
>> > > files('block.c'))
>> > >   softmmu_ss.add(when: zstd, if_true: files('multifd-zstd.c'))
>> > > -specific_ss.add(when: 'CONFIG_SOFTMMU', if_true:
>> > > files('dirtyrate.c', 'ram.c'))
>> > > +specific_ss.add(when: 'CONFIG_SOFTMMU',
>> > > +                if_true: files('dirtyrate.c', 'ram.c', 'special.c'))
>> > 
>> > Why not simply name this migration/vfio.c? That way we do not start
>> > mixed bag of everything target specific.
>> 
>> I don't mind ... well, if we have other small functions there in the future
>> that depend on CONFIG switches, a mixed bag file might not be such a bad
>> idea instead of having lots and lots of small other files ... OTOH, if there
>> is more vfio migration code in the works that might fit here, a name like
>> vfio.c would be better, of course. What do the maintainers think?
>
> Could this be done with stubs instead of an ifdef; i.e. a stub of
> 'vfio_mig_active' and 'vfio_mig_bytes_transferred'?

My understanding is that they can't (at least easily).
Because they are really target specific :-(

> As for naming 'special' is too generic.
> 'vfio' is too specific (especially since most vfio code ends up under
> hw/vfio)
>
> how about migration/target.c  as something which is explicit about why
> it's done that way.

I agree with the target name.

But I can't think of a really much easier way that what is in this
patch.

To make stubs you need to ifdef half of hw/vfio/vfio-common.h, or just
put the #ifdef in migration/target.c

Even althought I hate #ifdefs, I am not sure that the stubs options is
much clearly here.

Later, Juan.