[libvirt] The weird '@' in bind address on OSX

[Justin Clift]

Hi all,

As a general thought, would anyone be interested in getting this weird '@'
thing in the bind address on OSX fixed?

It shows up in both libvirtd and virsh:

  $ libvirtd 
  2012-02-08 08:39:53.211+: -1: info : libvirt version: 0.9.10
  2012-02-08 08:39:53.211+: -1: error : virNetSocketNewListenUNIX:346 : 
Failed to bind socket to '@/Users/jc/.libvirt/libvirt-sock': No such file or 
directory

and

  $ virsh
  Welcome to virsh, the virtualization interactive terminal.

  Type:  'help' for help with commands
 'quit' to quit

  virsh # version
  error: Failed to reconnect to the hypervisor
  error: no valid connection
  error: Failed to connect socket to '@/Users/jc/.libvirt/libvirt-sock': No 
such file or directory

Matthias mentioned it's due to using anonymous unix sockets, which
are only present on Linux:

  https://www.redhat.com/archives/libvirt-users/2011-November/msg00018.html

Guess that means libvirtd doesn't work on *BSD either?

Perhaps gnulib has a cross platform way to take care of this?

(Note, I could have sworn eblake started looking into this ages ago,
but then I moved on to the Aeolus team.  Unsure if there was useful
progress back then. ?)

Regards and best wishes,

Justin Clift

--
Aeolus Community Manager
http://www.aeolusproject.org


--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] The weird '@' in bind address on OSX

[Eric Blake]

On 02/08/2012 01:58 AM, Justin Clift wrote:
 Hi all,
 
 As a general thought, would anyone be interested in getting this weird '@'
 thing in the bind address on OSX fixed?

The @ is merely a visual place-holder for the actual NUL byte being
written into the actual socket structure handed to the kernel.  But you
are correct that:

 
 Matthias mentioned it's due to using anonymous unix sockets, which
 are only present on Linux:
 
   https://www.redhat.com/archives/libvirt-users/2011-November/msg00018.html

Use of a NUL byte in the first byte of a socket address is a Linux-only
hack to get an anonymous socket namespace.

 
 Guess that means libvirtd doesn't work on *BSD either?

Correct, for anywhere that libvirt currently uses an anonymous socket
(such as in qemu:///session).

 
 Perhaps gnulib has a cross platform way to take care of this?

No.  There's no way to fake anonymous namespace sockets if the kernel
does not have them.  Rather, libvirt needs to have a graceful fallback
to something in the file system (perhaps under ~/.libvirt/...) on
platforms that don't have an anonymous socket namespace.

 
 (Note, I could have sworn eblake started looking into this ages ago,
 but then I moved on to the Aeolus team.  Unsure if there was useful
 progress back then. ?)

No, I haven't really done anything with this.  Sorry.

-- 
Eric Blake   ebl...@redhat.com+1-919-301-3266
Libvirt virtualization library http://libvirt.org



signature.asc
Description: OpenPGP digital signature
--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list