Re: [virtio-dev] [PATCH] content: reserve device ID for virtio-pmem devices

[Stefan Hajnoczi <stefanha@redhat.com>]

On Wed, May 23, 2018 at 01:01:20AM -0400, Pankaj Gupta wrote:
> 
> Hi Stefan,
> 
> > >> We need a device ID for virtio-pmem devices. As 24 is already
> > >> requested by virtio-mem, so requesting next available(25)
> > >>
> > >> Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
> > >> ---
> > >>  content.tex | 2 ++
> > >>  1 file changed, 2 insertions(+)
> > > 
> > > virtio-mem is designed for hotplug.  Does it make sense to merge
> > > virtio-pmem into virtio-mem since you'll need hotplug too?
> > > 
> > > Stefan
> > > 
> > Hi Stefan,
> > 
> > Initially I had (and even proposed during some discussion) the same
> > idea. But it turns out that they are fundamentally different.
> > 
> > They are memory devices, yes, but that's the point where they stop being
> > similar.
> > 
> > While virtio-pmem wants to expose a memory region as persistent memory
> > and adding a flushing interface, virtio-mem is all about managing
> > (add/remove/online/offline) sub-chunks and applying protection
> > mechanisms on the hypervisor side to make sure unplugged memory cannot
> > (or only in some limited sense) be accessed.
> 
> Yes, we discussed possibility of merging both these devices in the past.
> As rightly explained by David, both these devices solving different problems.
> Also, their architectures are different, common part is both use memory device 
> & VIRTIO.
> 
> Also, as we plan to add more features specific to each virtio{mem/pmem}, it 
> will be more confusing and difficult to maintain. I think it makes more sense to 
> keep both these devices separate. 

Okay.

Stefan

Attachment: signature.asc
Description: PGP signature