Re: [Xen-devel] Xen + futexes

[Dario Faggioli <dario.faggioli@xxxxxxxxxx>]

On Fri, 2015-07-03 at 18:39 +0300, Vitaly Chernooky wrote:


> On Thu, Jul 2, 2015 at 1:48 PM, Dario Faggioli
> <dario.faggioli@xxxxxxxxxx> wrote:

>         Actually, we don't even have mutexes _within_ the hypervisor,
>         so I
>         struggle a bit to figure out how it would be useful to offer a
>         similar
>         support to whatever is userspace for us (i.e., what futexes
>         do, as far
>         as I can recall).
> 
> 
> Yes, it is reasonable because mutexes are enough hi-level stuff. But
> futexes is not so hi-level ...
>
Which is (one of) the reason(s) why they're so tricky. Anyway, I'm not
in principle opposed to this, especially not for this reason...
> 
>         So, that's why I was joking/saying about this being too much
>         theoretical: do you have a use case in mind where something
>         like that
>         can help and be useful? If yes, I'd be interested in hearing
>         about it.

> As far as we can observe sometimes in this mailing list we get RFCs
> for new intercommunication mechanisms. If i'm not mistaken last one
> was Xen Sockets. It looks like existing mechanism are not enough
> useful for many people. It is the reason why I started to discuss this
> ideas.
> 
Well, of course it's fine to toss new ideas and send out RFCs about
(potential) new stuff!

However, it might be me, but I think that, when proposing something like
this, it would be useful to have in mind (and describe it to us) at
least a potential use case, and/or a situation where such a new
mechanism could be used fruitfully, make things easier, improve
performance... I don't know... something! :-)

I know, it's a bit of a chicken-&-egg: if it's not there, there's no use
case for it. But at least in a speculative manner, having something in
mind is important, IMO. And not only to show the need for something, but
also for having a chance to get the implementation and --most
important-- the interface, right (well, that's almost impossible...
let's say at least *not completely wrong* :-D).

I think Xen's complexity is beyond the point where we add stuff 'just'
for the sake of adding them, or because they're (may be?) cool, and then
we'll see what to do with them afterwards (but that's only my personal
opinion).

Also, I'm not saying that this thing I'm asking is a prerequisite of
some sort for proposing adding a feature... It's just that you asked
what our thoughts were about adding futexes, and my own thoughts are
"what for?"

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: signature.asc
Description: This is a digitally signed message part

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