On Fri, 3 Jul 2020 10:06:12 +0800 Wei Yang <[email protected]> wrote:
> Currently we have following call flow: > > tracer_init_tracefs() > tracing_init_dentry() > event_trace_init() > tracing_init_dentry() > > This shows tracing_init_dentry() is called twice in this flow and this > is not necessary. There's no reason to have patch 4 and 5 separate. Fold the two together. If you want, you can create another patch that changes tracing_init_dentry() to return a integer, as you point out, it never returns an actual dentry. No reason for having it return a pointer then. -- Steve > > Let's remove the second one when it is for sure be properly initialized. > > Signed-off-by: Wei Yang <[email protected]> > --- > kernel/trace/trace_events.c | 5 ----- > 1 file changed, 5 deletions(-) > > diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c > index 8b3aa57dcea6..76879b29cf33 100644 > --- a/kernel/trace/trace_events.c > +++ b/kernel/trace/trace_events.c > @@ -3434,7 +3434,6 @@ early_initcall(event_trace_enable_again); > __init int event_trace_init(void) > { > struct trace_array *tr; > - struct dentry *d_tracer; > struct dentry *entry; > int ret; > > @@ -3442,10 +3441,6 @@ __init int event_trace_init(void) > if (!tr) > return -ENODEV; > > - d_tracer = tracing_init_dentry(); > - if (IS_ERR(d_tracer)) > - return 0; > - > entry = tracefs_create_file("available_events", 0444, NULL, > tr, &ftrace_avail_fops); > if (!entry)
