On Fri, Mar 09, 2018 at 12:56:12 +0100, Michal Privoznik wrote:
> https://bugzilla.redhat.com/show_bug.cgi?id=1553162
> When validating a device XML config we check if user provided
> alias is unique. We do this by maintaining a hash table of device
> aliases as we iterated over all devices defined for the domain.
> However, it may happen that what appears as two devices in domain
> XML is in fact just one interface in hypervisor. For instance in
> qemu driver this is true for uhci/ehci controllers. In that case
> an error is reported even though it is not actually an error. At
> any rate, we can assume libvirt generated aliases to be unique
> and thus really check user provided ones only.

So I find some parts of this description slightly misleading. I agree
that we should only validate that the user-provided aliases are unique,
since libvirt should do the proper thing otherwise and the user-aliases
are in it's own namespace.

I think though that the hint to the usb controller requiring multiple
devices from the libvirt view map to a single device in qemu, or any
other reason why two libvirt entries should have the same alias is not
future-proof enough. With this commit you make it possible to use it
with the default alias, but it still might be possible to use with a
user alias.

ACK to this patch, since I think it's the correct thing to do, but
please delete the irrelevant blurb about uhci/ehci in qemu.

Attachment: signature.asc
Description: PGP signature

libvir-list mailing list

Reply via email to