From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Daniel P. Berrange" <berrange@redhat.com>
Subject: Re: [systemd-devel] udev virtio by-path naming
Date: Tue, 28 Feb 2017 19:28:51 +0000
Message-ID: <20170228192851.GC10067__21170.0470456437$1488310145$gmane$org@redhat.com>
References: <de55b0f4-4582-cf06-4b0d-12c2282406a8@linux.vnet.ibm.com>
	<20170220151432.GA15888@gardel-login>
Reply-To: "Daniel P. Berrange" <berrange@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <virtualization-bounces@lists.linux-foundation.org>
Content-Disposition: inline
In-Reply-To: <20170220151432.GA15888@gardel-login>
List-Unsubscribe: <https://lists.linuxfoundation.org/mailman/options/virtualization>,
	<mailto:virtualization-request@lists.linux-foundation.org?subject=unsubscribe>
List-Archive: <http://lists.linuxfoundation.org/pipermail/virtualization/>
List-Post: <mailto:virtualization@lists.linux-foundation.org>
List-Help: <mailto:virtualization-request@lists.linux-foundation.org?subject=help>
List-Subscribe: <https://lists.linuxfoundation.org/mailman/listinfo/virtualization>,
	<mailto:virtualization-request@lists.linux-foundation.org?subject=subscribe>
Sender: virtualization-bounces@lists.linux-foundation.org
Errors-To: virtualization-bounces@lists.linux-foundation.org
To: Lennart Poettering <lennart@poettering.net>
Cc: virtualization@lists.linux-foundation.org, systemd-devel@lists.freedesktop.org
List-Id: virtualization@lists.linuxfoundation.org

On Mon, Feb 20, 2017 at 04:14:32PM +0100, Lennart Poettering wrote:
> On Mon, 20.02.17 15:34, Viktor Mihajlovski (mihajlov@linux.vnet.ibm.com) wrote:
> 
> > But then, I find this naming scheme somewhat weird.
> > A virtio disk shows up as a regular PCI function on the PCI
> > bus side by side with other (non-virtio) devices. The naming otoh
> > suggests that virtio-pci is a subsystem of its own, which is simply
> > incorrect from a by-path perspective.
> > 
> > Using just the plain PCI path id is actually sufficient to identify
> > a virtio disk by its path. This would be in line with virtio
> > network interface path names which use the plain PCI naming.
> > 
> > One could argue about back-level compatibility, but virtio by-path
> > naming has changed multiple times. We have seen virtio-pci-virtio<n>
> > (not predictable), pci-<busid> and virtio-pci-<busid> already. It
> > might be a good time now to settle on a common approach for all
> > virtio types.
> > 
> > For the reasons above, I'd vote for <subsystem>-<busid>, which
> > would work for PCI and CCW, not sure about ARM MMIO though.
> > Opinions?

Virtio MMIO devices are identified by a unique control register
base address. eg 0x3000. So I think <subsystem>-<busid> would
work fine to all cases PCI, CCW & MMIO.  Certainly it is moire
correct than hardcoding virtio-pci as a prefix - that's just
plain broken for non-PCI transports.

> So, to make this clear, we in systemd are kinda interested in
> splitting out these virtio helpers into some external project
> maintained by virtio peopl. We as systemd/udev maintainers have very
> little understanding of the underlying technology, so we can't really
> be any good maintainers of this, and we can't really comment on this
> stuff, in particular when it gets more exotic, like the CCW stuff.
> 
> Even better would be if the kernel would do the naming on its own, and
> maybe just provide us with a sysattr on the relevant devices that we
> can read to determine the path from, so that we don#t have to maintain
> this at all in userspace. That way, the driver folks on the kernel
> side can use any naming they like without ever having to patch this
> into systemd or udev.
> 
> This is similar to SCSI stuff and all things like that: the more
> exotic it gets the less place this really has in systemd, we are not
> the right maintainers for this. And given that this is all nicely
> pluggable (you can ship your own udev extensions externally very
> easily), there's really no reason for this to be in systemd/udev.

The other post about ptp-kvm rules reminded me that I wanted to
respond to this mail too.

The problem with splitting these rules out into a separate project
is that there's no other existing place that they would live. The
"virtio people" as a group merely write specifications. The actual
implementation of those specs is done by multiple other independant
groups - QEMU (for host side, though other host side impls exist
too) and Linux (for guest side). The udev rules are Linux guest
support pieces, but of course Linux itself doesn't distribute udev
rules - it delegated that job to the udev package hence why they
are here currently. So I don't see that pushing the rules out of
the udev repo would be beneficial to people building VMs.

> Anyway, I fear you're going to have a hard time involving us in a
> technical discussions about the issue you are raising, since quite
> frankly we have no clue about virtio...

Could it be as simple as having a couple of people nominated as the
technical point of contact for the virtio rules, who can be CC'd
to get answers any questions that may need answering ? I don't have
time to actively monitor systemd pull requests for changes affecting
virtio, but I'd be ok with being pinged if issues come up that need
assistance & can pull in other virt experts where needed.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://entangle-photo.org       -o-    http://search.cpan.org/~danberr/ :|