Re: [PATCH 14/15] null-machine: do not create a default memdev

[Paolo Bonzini]

On 14/12/20 12:53, Igor Mammedov wrote:
> On Sat, 12 Dec 2020 00:24:25 +0100
> Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> > On 07/12/20 17:43, Igor Mammedov wrote:
> > > >        mc->default_ram_size = 0;
> > > > -    mc->default_ram_id = "ram";
> > > > +    mc->default_ram_id = NULL;
> > > probably that will break:
> > >
> > >    QEMU -m X -M none
> >
> > No, it works.  "-m" is simply ignored, because the default memdev is
> > created here:
> you mean it doesn't explode.
> By default with current code memdev should not be created,
> but it is created if users asks for it with -m.

Oh ok, got it.  I'll leave this patch aside for now and later, when I 
get round to posting the ram_memdev_id cleanup, try to remember why I 
added it.

> I'm not sure that any RAM on null machine is of any use at all
> but the if we decide to get rid of it completely, then we need
> to clean up
>      /* RAM at address zero */
>      if (mch->ram) {
>          memory_region_add_subregion(get_system_memory(), 0, mch->ram);
>      }

This would still be needed because you _can_ get a non-null mch->ram 
(via "-M memdev") even if mc->default_ram_id == NULL.

Paolo

> and commit title/message should match what patch does and why.
>
>
> >       if (machine_class->default_ram_id && current_machine->ram_size &&
> >           numa_uses_legacy_mem() && !current_machine->ram_memdev_id) {
> >           create_default_memdev(current_machine, mem_path);
> >       }
> >
> > and is thus skipped for -M none.
> >
> > Paolo
> >
> > > maybe there is  a bug over there but
> > >       "mc->default_ram_size = 0"
> > > above, should result in
> > >       current_machine->ram_size == 0
> > > in case user hasn't provided "-m"
> > > and hence memdev shouldn't be created
> > >