[DTrace-devel] [PATCH 2/4] fbt: clean up fprobe/kprobe support

[Kris Van Hees]

On Thu, Dec 05, 2024 at 05:58:04PM -0500, Eugene Loh wrote:
> On 12/5/24 17:17, Kris Van Hees wrote:
> 
> > On Thu, Dec 05, 2024 at 05:05:15PM -0500, Eugene Loh wrote:
> > > I wouldn't mind a few commit msg words on what "clean up" means.
> > > 
> > > Mostly, how/where is dt_fbt_fprobe used?  Put another way...
> > > 
> > > On 12/5/24 13:53, Kris Van Hees via DTrace-devel wrote:
> > > > diff --git a/libdtrace/dt_prov_fbt.c b/libdtrace/dt_prov_fbt.c
> > > > @@ -73,9 +75,10 @@ static int populate(dtrace_hdl_t *dtp)
> > > >    	dtrace_syminfo_t	sip;
> > > >    	dtrace_probedesc_t	pd;
> > > > -	impl = BPF_HAS(dtp, BPF_FEAT_FENTRY) ? &dt_fbt_fprobe : &dt_fbt_kprobe;
> > > > +	if (!BPF_HAS(dtp, BPF_FEAT_FENTRY))
> > > > +		dt_fbt = dt_fbt_kprobe;
> > > If the BPF_HAS() test passes, shouldn't we have basically
> > > dt_fbt=dt_fbt_fprobe?  In which case, one might arguably go back to the
> > > ternary op?
> > Since dt_fbt is initialized as dt_fbt_fprobe (the default), this conditional
> > sets it to dt_fbt_kprobe if fprobes are not available.  I figured that was
> > obvious from the code.
> 
> Sorry, where should I look to see that dt_fbt is initialized to
> dt_fbt_fprobe?

Oh right - that was in the previous version.  But on OL7, gcc complained about
that assignment.  But I didn't complete the change in the patch I posted.

Let me post the corrct one (and yes, the ternary comes back).