Re: [Qemu-devel] [PATCH v2 1/2] trace: add TRACE_<event>_BACKEND_DSTATE()

[Stefan Hajnoczi]

On Mon, Jul 31, 2017 at 04:16:39PM +0100, Daniel P. Berrange wrote:
> On Mon, Jul 31, 2017 at 03:07:17PM +0100, Stefan Hajnoczi wrote:
> > diff --git a/scripts/tracetool/backend/dtrace.py 
> > b/scripts/tracetool/backend/dtrace.py
> > index c6812b70a2..17f902cc62 100644
> > --- a/scripts/tracetool/backend/dtrace.py
> > +++ b/scripts/tracetool/backend/dtrace.py
> > @@ -44,8 +44,20 @@ def generate_h_begin(events, group):
> >      out('#include "%s"' % header,
> >          '')
> >  
> > +    # SystemTap defines <provider>_<name>_ENABLED() but other DTrace
> > +    # implementations might not.
> > +    for e in events:
> > +        out('#ifndef QEMU_%(uppername)s_ENABLED',
> > +            '#define QEMU_%(uppername)s_ENABLED() false',
> > +            '#endif',
> > +            uppername=e.name.upper())
> 
> 
> IIUC, this means that on other DTrace impls, any trace points guarded by
> QEMU_*_ENABLED() will never run, even if DTrace probes are set. Having
> some trace points silently never run makes it pretty useless IMHO.
> 
> IOW, you need the opposite, #define it to true. The probe will still
> only be executed if DTrace has enabled it, but you'll have to take the
> hit of evaluating the probe arguments regardless. Not as optimized
> performance-wise, but functionally correct at least.

Good point!  Thank you, will fix when merging.

Stefan

signature.asc
Description: PGP signature