Re: [PATCH v2 42/50] lasi: move from hw/hppa to hw/misc

[Mark Cave-Ayland]

On 08/05/2022 04:07, Richard Henderson wrote:

> On 5/4/22 04:25, Mark Cave-Ayland wrote:
> > Move the LASI device implementation from hw/hppa to hw/misc so that it is
> > located with all the other miscellaneous devices.
> >
> > Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
> > Acked-by: Helge Deller <deller@gmx.de>
> > ---
> >   MAINTAINERS                         | 2 ++
> >   hw/hppa/Kconfig                     | 1 +
> >   hw/hppa/machine.c                   | 2 +-
> >   hw/hppa/meson.build                 | 2 +-
> >   hw/hppa/trace-events                | 5 -----
> >   hw/misc/Kconfig                     | 3 +++
> >   hw/{hppa => misc}/lasi.c            | 3 +--
> >   hw/misc/meson.build                 | 3 +++
> >   hw/misc/trace-events                | 5 +++++
> >   {hw/hppa => include/hw/misc}/lasi.h | 0
> >   10 files changed, 17 insertions(+), 9 deletions(-)
> >   rename hw/{hppa => misc}/lasi.c (99%)
> >   rename {hw/hppa => include/hw/misc}/lasi.h (100%)
>
> I don't understand hw/misc, or why this is a better categorization than hw/hppa.

This one is a bit of a grey area at the moment. There have been some discussions 
around having a hard separation between boards and devices, so that devices live in 
one part of the directory hierarchy and boards live in another. I think possibly 
Markus was advocating but this, but of course that would involve quite some work.

The reason for leaning in this direction is that by having the device separate from 
the board then it requires uses of QOM APIs in a way that don't break this 
abstraction, which I think makes things more future-proof.

There are already quite a few machine-specific devices throughout the tree: good 
examples of this would be hw/pci-host and hw/misc where there are a number of devices 
that it doesn't make sense to wire up to anything except one particular machine, so 
we're certainly not doing anything new here.


ATB,

Mark.