Re: [PATCH v3 3/4] libxl: add rtds scheduler

[Dario Faggioli <dario.faggioli@citrix.com>]

[-- Attachment #1.1: Type: text/plain, Size: 2231 bytes --]

On mar, 2014-09-16 at 17:54 +0100, Ian Campbell wrote:
> On Tue, 2014-09-16 at 09:27 +0200, Dario Faggioli wrote:
> > On Mon, 2014-09-15 at 23:32 -0400, Meng Xu wrote:
> > 
> > > 2014-09-15 21:49 GMT-04:00 Ian Campbell <ian.campbell@citrix.com>:
> > 
> > >         Is a 0.5hr maximum period considered to be larger than any
> > >         practically
> > >         useful value?
> > > 
> > > 
> > > I think so. 
> > >
> > Yes, it is. Most of the people using this scheduler will be interested
> > in "a few [decades, hundreds] millisecond" period and budget. Someone
> > might want to try microsecs, someone may be fine with seconds.
> 
> OK, so no need for me to ask if it should be a 64bit value then.
> 
It actually was. Then Jan asked whether it really needed to be, and we
all agreed it did not.

TBF, what we were talking in tha thread was the xc and hv interface.
Nothing forbids the two interfaces to differ in the types of this field,
I think. I guess the question here is whether we would like, at the
libxl level, to be a bit, let's say, 'less strict'.

That would mean, e.g., using something like int64 or uint64 in libxl
(type declarations and all the related functions), and then do some
range checking, in the libxl implementation, before passing down the
values to libxc. That would be because, while libxc and Xen interface
can change, libxl should not. One possible reason could be, for example,
if we at some point will want for the budget and period to be specified
in ns, rather than us.

Honestly, I don't foresee such a thing becoming any useful, not even in
mid or long term future. It is what at least Linux sort of does (POSIX's
timespec-s are used, there), though, so it at least is worth some
thinking?

As I said already, I think this is a fine interface, even with stability
in mind, but I'd be fine with a int64_t vesion, here in libxl, if people
feels it may be better. :-)

Regards,
Dario

-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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