Re: [Xen-devel] [PATCH 16/19] libxl: suspend: Abolish usleeps in domain suspend wait

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Thu, 2014-03-13 at 18:29 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [PATCH 16/19] libxl: suspend: Abolish usleeps in 
> domain suspend wait"):
> > On Tue, 2014-03-04 at 14:56 +0000, Ian Jackson wrote:
> > > Replace the use of a loop with usleep().
> ...
> > > +    rc = libxl__ev_xswatch_register(gc, &dss->guest_watch,
> > > +                                    suspend_common_wait_guest_watch,
> > > +                                    "@releaseDomain");
> > > +    if (rc) goto err;
> > > +
> > > +    rc = libxl__ev_time_register_rel(gc, &dss->guest_timeout, 
> > > +                                     suspend_common_wait_guest_timeout,
> > > +                                     60*1000);
> > 
> > This smells a lot like the xswait stuff you introduced towards the
> > beginning of the series.
> > 
> > Is the difference just that you cannot read the "path" @releaseDomain?
> 
> Precisely.
> 
> > If so can we paper over that somehow in the helper?
> 
> Not without complicating its interface, I think.  We would really want
> to prevent the @releaseDomain "path" being read at all.
> 
> I guess we could special-case paths starting with "@", the way
> xenstore does, and always return ENOENT.

I wonder what xs_read on @releaseDomain actually does -- I'd not be
surprised if it was ENOENT. I wonder even more what xs_write on
@releaseDomain would do, hopefully "bugger off" but you never know.

>   I'm not sure whether this
> wrinkle (which would end up in the docs for the general facility) is
> worth the saving in the one call site where it would be relevant.

I'm not sure either, I'll leave it to your best judgement. If you want
to stick with this variant then:
Acked-by: Ian Campbell <ian.campbell@xxxxxxxxxx>

Ian.


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