Re: [Xen-devel] [RFC] PVFB: Add refresh period to XenStore parameters?

[Samuel Thibault <samuel.thibault@xxxxxxxxxxxxx>]

Markus Armbruster, le Fri 09 May 2008 10:43:15 +0200, a écrit :
> > +    /* Will have to be disabled for frontends without feature-update */
> 
> I think I asked you to put this comment here, but is it still correct
> for current XENFB_TYPE_REFRESH_PERIOD semantics?

Considering the clarification of the semantic, it can be dropped indeed.

> > + *
> > + * If the frontend uses the advice, it should refresh and send an update 
> > event
> > + * in response to this event.
> >   */
> 
> You delete the bit about ignoring unknown events.  Oops.

Oops.

> What about this:

Ok.

> > +#define XENFB_TYPE_REFRESH_PERIOD 1
> > +
> > +struct xenfb_refresh_period
> > +{
> > +    uint8_t type;    /* XENFB_TYPE_UPDATE_PERIOD */
> > +    uint32_t period; /* period of refresh, in ms, 0 if no refresh is 
> > needed */
> > +};
> >  
> >  #define XENFB_IN_EVENT_SIZE 40
> >  
> >  union xenfb_in_event
> >  {
> >      uint8_t type;
> > +    struct xenfb_refresh_period refresh_period;
> 
> Time unit?

There is one in the structure above.

> I'd be tempted to use a frequency instead of a period, just because
> that doesn't require a special value for "no updates".  Strictly a
> matter of taste.

The "problem" of an integer frequency is that it does not permit a
period of more than one second, which may become a limit in some odd
situations (thousands of VMs waking every second?)

Samuel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel