Re: [Xen-devel] [RFC PATCH 2/6] libxl: stop using libxl__xs_mkdir() for ~/control/shutdown

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [RFC PATCH 2/6] libxl: stop using libxl__xs_mkdir() 
for ~/control/shutdown"):
> On Tue, 2015-11-24 at 17:20 +0000, Paul Durrant wrote:
> > The issue, as I said, is the initial state of the node. If you use
> > XS_MKDIR then it is not guaranteed to be empty.
> 
> Just to satisfy my curiosity, how can it be non-empty? What else could it
> possibly contain, just garbage?

AIUI Paul is worried about the hypothetical possibility that someone
might call libxl__xs_mkdir
  - on a path which already exists;
  - and has a nonempty value;
  - expecting the nonempty value to be cleared.

I think this is too remote a possibility to be concerned about.  But
since the WRITE operation has identical semantics to MKDIR _except_
that it overwrites the existing value, I don't mind if Paul cares
about this and wants to change libxl to always use WRITE and never
MKDIR.[1]

Ian.

[1] There are AFAIAA no existing xenstore protocols or uses where a
node is deliberately used both to store a value, and as a parent for
descendent nodes.  We don't seem likely to introduce any such thing,
because it would be confusing.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel