Re: [PATCH v3 07/16] hw/i386/vmport: Introduce vmport.h

[Liran Alon]

On 14/03/2020 10:31, Philippe Mathieu-Daudé wrote:
> On 3/13/20 11:38 PM, Liran Alon wrote:
> > On 13/03/2020 21:57, Philippe Mathieu-Daudé wrote:
> > > On 3/12/20 5:54 PM, Liran Alon wrote:
> > > > No functional change. This is mere refactoring.
> > > >
> > > > Suggested-by: Michael S. Tsirkin <address@hidden>
> > > > Signed-off-by: Liran Alon <address@hidden>
> > > > ---
> > > >   hw/i386/pc.c             |  1 +
> > > >   hw/i386/vmmouse.c        |  1 +
> > > >   hw/i386/vmport.c         |  1 +
> > > >   include/hw/i386/pc.h     | 13 -------------
> > > >   include/hw/i386/vmport.h | 16 ++++++++++++++++
> > >
> > > What about moving it to hw/i386/vmport.h (no under include/)?
> > >
> > > Reviewed-by: Philippe Mathieu-Daudé <address@hidden>
> > Can you explain the logic that separates between hw/i386/*.h to 
> > include/hw/i386/*.h?
>
> Headers in the include/hw/ namespace can be consumed by all machine 
> targets.
But this doesn't seem true for headers in include/hw/i386/*.h...
It contains things that are target-specific. E.g. ioapic.h, x86-iommu.h, 
intel_iommu.h and etc.
I still don't quite understand the separation between these directories. 
It seems both are i386-specific and one of them shouldn't exists.
> If this is a target-specific device, having it local to the target 
> (hw/i386/) protect generic code (and other targets) of using it. This 
> helps detecting wrong dependencies between components.
>
> > If it makes sense, sure I will move it. I just don't know what is the 
> > convention here.
>
> Michael/Paolo/Eduardo what do you recommend?