RE: Libvirt NVME support

-----Original Message----- From: Peter Krempa <pkrempa(a)redhat.com&gt; Sent: 23 November 2020 16:56 To: Thanos Makatos <thanos.makatos(a)nutanix.com&gt; Cc: Suraj Kasi <suraj.kasi(a)nutanix.com>; libvirt-list(a)redhat.com; John Levon <john.levon(a)nutanix.com&gt; Subject: Re: Libvirt NVME support On Mon, Nov 23, 2020 at 16:48:55 +0000, Thanos Makatos wrote: > > On Mon, Nov 23, 2020 at 13:07:51 +0000, Thanos Makatos wrote: > > > > > > > On Mon, Nov 23, 2020 at 09:47:23 +0000, Thanos Makatos wrote: > > > > > > On Thu, Nov 19, 2020 at 10:17:56 +0000, Thanos Makatos wrote: > > > > > > > > > > Revistiting your initial suggestion, it should be something like this > > > > > (s/sdb/nvme0): > > > > > > > > > > <disk type='file' device='disk'> > > > > > <source dev='/Host/QEMUGuest2.qcow2'/> > > > > > <target dev='nvme0' bus='nvme'/> > > > > > <address type='drive' controller='1' bus='0' target='0' unit='1'/> > > > > > </disk> > > > > > > > > Note that the parser for 'dev' is a bit quirky, old, and used in many > > > > places besides the qemu driver. It's also used with numbers in non- qemu > > > > cases. Extending that to parse numbers for nvme but not for sda might > > > > become ugly very quickly. Sticking with a letter at the end ('nvmea' > > > > might be a more straightforward approach. > > > > > > Then I think we should just stick with 'nvme'. > > > > You still need a way to "index" it somehow. The target must be unique > > for each disk. > > I think I've misunderstood something, I thought controller='1' in <address ...> > refers to index='1' in <controller ...>. So <address ...> should be: > > <address type='nvme' index='1' controller='1' ns='2'/> > > What's controller='1' then? What I meant by the above is that the value of "<target dev='THIS'" must be unique for every <disk>. I also wanted to advice to not use numbers for making it unique. Numbers used for it have a legacy meaning. Your suggested <address type='nvme' design looks good.