Re: [PATCH 1/2] tracing: Remove "__attribute__()" from the type field of event format

[Masami Hiramatsu (Google) <mhiramat@kernel.org>]

On Fri, 11 Jul 2025 12:03:22 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Fri, 11 Jul 2025 14:37:03 +0900
> Masami Hiramatsu (Google) <mhiramat@kernel.org> wrote:
> 
> > I investigated this but it is not possible to use update_event_fields()
> > because that function is only used if the 
> > CONFIG_TRACE_EVAL_MAP_FILE=y.
> 
> That had better *not* be true, otherwise all the TRACE_DEFINE_ENUM() in
> include/trace/events/*.h would be useless and those events would not
> parse. Note, I usually run without that config enabled, so it would
> most definitely break on me if this was true.
> 
> In the code we have:
> 
> ----------------------8<----------------------
> #else /* CONFIG_TRACE_EVAL_MAP_FILE */
> static inline void trace_create_eval_file(struct dentry *d_tracer) { }
> static inline void trace_insert_eval_map_file(struct module *mod,
> 			      struct trace_eval_map **start, int len) { }
> #endif /* !CONFIG_TRACE_EVAL_MAP_FILE */
> 
> static void trace_insert_eval_map(struct module *mod,
> 				  struct trace_eval_map **start, int len)
> {
> 	struct trace_eval_map **map;
> 
> 	if (len <= 0)
> 		return;
> 
> 	map = start;
> 
> 	trace_event_eval_update(map, len);
> 
> 	trace_insert_eval_map_file(mod, start, len);
> }
> ---------------------->8----------------------
> 
> Notice the "#endif". The trace_insert_eval_map_file() is a nop, but the
> trace_event_eval_update() is not. That has the call to
> update_event_printk() and update_event_fields().
> 
> So it can most definitely be used when that config is not defined. That
> config only creates a file to show you *what* was replaced. It doesn't
> stop the replacing.

Hmm, Ok. But when I sanitized the field->type in
update_event_fields(), it did not work. So something
we missed.

Thanksm

-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>