Re: [Qemu-devel] [PATCH v5 5/6] Add memfd based hostmem

["Marc-André Lureau" <marcandre.lureau@redhat.com>]

----- Original Message -----
> On Mon, Oct 23, 2017 at 10:59:09AM +0100, Marc-André Lureau wrote:
> > Add a new memory backend, similar to hostmem-file, except that it
> > doesn't need to create files. It also enforces memory sealing.
> > 
> > This backend is mainly useful for sharing the memory with other
> > processes.
> > 
> > Note that Linux supports transparent huge-pages of shmem/memfd memory
> > since 4.8. It is relatively easier to set up THP than a dedicate
> > hugepage mount point by using "madvise" in
> > /sys/kernel/mm/transparent_hugepage/shmem_enabled.
> > 
> > Since 4.14, memfd allows to set hugetlb requirement explicitly.
> > 
> > Usage:
> > -object memory-backend-memfd,id=mem1,size=1G
> 
> [snip]
> 
> > +static void
> > +memfd_backend_class_init(ObjectClass *oc, void *data)
> > +{
> > +    HostMemoryBackendClass *bc = MEMORY_BACKEND_CLASS(oc);
> > +
> > +    bc->alloc = memfd_backend_memory_alloc;
> > +
> > +    object_class_property_add_bool(oc, "hugetlb",
> > +                                   memfd_backend_get_hugetlb,
> > +                                   memfd_backend_set_hugetlb,
> > +                                   &error_abort);
> 
> I tend to think that instead of a bool hugetlb, we should take an
> integer page size instead eg  hugepagesize=2M instead of hugetlb=true
> 

Well, how would you then create a memfd without explicit hugetlb request or one with default hugetlb size?

I think size should be a different property. I'll add it.

> 
> > diff --git a/qemu-options.hx b/qemu-options.hx
> > index 3728e9b4dd..5828caefeb 100644
> > --- a/qemu-options.hx
> > +++ b/qemu-options.hx
> > @@ -4208,6 +4208,21 @@ that @option{discard-data} is only an optimization,
> > and QEMU
> >  might not discard file contents if it aborts unexpectedly or is
> >  terminated using SIGKILL.
> >  
> > +@item -object
> > memory-backend-memfd,id=@var{id},size=@var{size},seal=@var{on|off},hugetlb=@var{on|off}
> > +
> > +Creates an anonymous memory file backend object, which allows QEMU to
> > +share the memory with an external process in some cases (e.g. when
> > +using vhost-user). The memory is allocated with memfd and optional
> > +sealing. (Linux only)
> > +
> > +The @option{id} parameter is a unique ID that will be used to
> > +reference this memory region when configuring the @option{-numa}
> > +argument. The @option{size} option provides the size of the memory
> > +region, and accepts common suffixes, eg @option{500M}. The
> > +@option{seal} option creates a sealed-file, that will block further
> > +resizing the memory ('on' by default). The @option{hugetlb} option
> > +specify the file to be created resides in the hugetlbfs filesystem.
> 
> This should document that you can't combine sealing and huge pages
> 

Ok, I am wondering if this is a temporary limitation, I have asked the patch author about it (it looks possible to me, I'll eventually try implementing it)