Re: [PATCH v1 7/9] memory: introduce RAM_NORESERVE and wire it up in qemu_ram_mmap()

[David Hildenbrand]

On 02.03.21 21:54, Peter Xu wrote:
> On Tue, Mar 02, 2021 at 08:02:34PM +0100, David Hildenbrand wrote:
> > > > @@ -174,12 +175,18 @@ void *qemu_ram_mmap(int fd,
> > > >                        size_t align,
> > > >                        bool readonly,
> > > >                        bool shared,
> > > > -                    bool is_pmem)
> > > > +                    bool is_pmem,
> > > > +                    bool noreserve)
> > >
> > > Maybe at some point we should use flags too here to cover all bools.
> >
> > Right. I guess the main point was to not reuse RAM_XXX.
> >
> > Should I introduce RAM_MMAP_XXX ?
>
> Maybe we can directly use MAP_*?  Since I see qemu_ram_mmap() should only exist

I think the issue is that there is for example no flag that corresponds 
to "is_pmem" - and the fallback logic in our mmap code to make "is_pmem" 
still work is a little bit more involved. In addition, "readonly" 
translates to PROT_READ ...

> with CONFIG_POSIX.  However indeed I see no sign to extend more bools in the
> near future either, so maybe also fine to keep it as is, as 4 bools still looks
> okay - your call. :)

Well, I had the same idea when I added yet another bool :) But I guess 
we won't be adding a lot of additional flags in the near future. 
(MAP_POPULATE? ;) fortunately we use a different approach to populate 
memory)

I'll think about it, not sure yet if this is worth proper flags. Thanks!

--
Thanks,

David / dhildenb