On Mon, Jan 23, 2017 at 7:34 PM, Daniel P. Berrange <berrange(a)redhat.com>
wrote:
On Mon, Jan 23, 2017 at 07:26:58PM +0530, Nitesh Konkar wrote:
> Hi Daniel,
>
>
> On Mon, Jan 23, 2017 at 4:08 PM, Daniel P. Berrange <berrange(a)redhat.com
>
> wrote:
>
> > On Mon, Jan 23, 2017 at 04:03:21PM +0530, Nitesh Konkar wrote:
> >
> > > @@ -9570,16 +9570,17 @@ qemuDomainSetPerfEvents(virDomainPtr dom,
> > > if (def) {
> > > for (i = 0; i < nparams; i++) {
> > > virTypedParameterPtr param = ¶ms[i];
> > > - enabled = param->value.b;
> > > + state = param->value.i;
> >
> > NACK, this semantic change is not backwards compatible - this will
break
> > all applications & language bindings currently using libvirt perf
events.
> >
> > Oh I see . I didn't realize that. Any other API/place you suggest,
through
> which
> we can get this functionality in?
Why not just reset counters to zero when first enabling the perf
event. If people want to reset them again later, they can just
disable+enable the event to get the reset
I think, if this patch gets in:
https://www.redhat.com/
archives/libvir-list/2017-January/msg00956.html
then disable+enable would be equivalent to a reset operation.
As of today, every enable request is creating a new fd, thus leaking the
previous one.
Thanks,
Nitesh