Re: [for-next][PATCH 7/8] tracing/probe: Reject exactly same probe event

[Masami Hiramatsu <mhiramat@kernel.org>]

On Mon, 23 Sep 2019 16:12:53 +0530
Srikar Dronamraju <srikar@linux.vnet.ibm.com> wrote:

> Hey Masami, Steven
> 
> >  
> > +static bool trace_kprobe_has_same_kprobe(struct trace_kprobe *orig,
> > +					 struct trace_kprobe *comp)
> > +{
> > +	struct trace_probe_event *tpe = orig->tp.event;
> > +	struct trace_probe *pos;
> > +	int i;
> > +
> > +	list_for_each_entry(pos, &tpe->probes, list) {
> > +		orig = container_of(pos, struct trace_kprobe, tp);
> > +		if (strcmp(trace_kprobe_symbol(orig),
> > +			   trace_kprobe_symbol(comp)) ||
> > +		    trace_kprobe_offset(orig) != trace_kprobe_offset(comp))
> > +			continue;
> > +
> > +		/*
> > +		 * trace_probe_compare_arg_type() ensured that nr_args and
> > +		 * each argument name and type are same. Let's compare comm.
> > +		 */
> > +		for (i = 0; i < orig->tp.nr_args; i++) {
> > +			if (strcmp(orig->tp.args[i].comm,
> > +				   comp->tp.args[i].comm))
> > +				continue;
> 
> In a nested loop, *continue* is going to continue iterating through the
> inner loop. In which case, continue is doing nothing here. I thought we
> should have used a goto instead. No?  To me, continue as a last statement of
> a for loop always looks weird.

Oops, thanks for pointing it out!

> 
> > +		}
> > +
> > +		return true;
> > +	}
> 
> I think we need something like this:
> 
> 	list_for_each_entry(pos, &tpe->probes, list) {
> 		orig = container_of(pos, struct trace_kprobe, tp);
> 		if (strcmp(trace_kprobe_symbol(orig),
> 			   trace_kprobe_symbol(comp)) ||
> 		    trace_kprobe_offset(orig) != trace_kprobe_offset(comp))
> 			continue;
> 
> 		/*
> 		 * trace_probe_compare_arg_type() ensured that nr_args and
> 		 * each argument name and type are same. Let's compare comm.
> 		 */
> 		for (i = 0; i < orig->tp.nr_args; i++) {
> 			if (strcmp(orig->tp.args[i].comm,
> 				   comp->tp.args[i].comm))
> 				goto outer_loop;
> 
> 		}
> 
> 		return true;
> outer_loop:
> 	}

Correct, that's what I intended.
Could you make a fix patch on top of it? (or do I?)

Thank you,

> 
> 
> > +
> > +	return false;
> > +}
> > +
> >  
> 
> ......
> 
> > +static bool trace_uprobe_has_same_uprobe(struct trace_uprobe *orig,
> > +					 struct trace_uprobe *comp)
> > +{
> > +	struct trace_probe_event *tpe = orig->tp.event;
> > +	struct trace_probe *pos;
> > +	struct inode *comp_inode = d_real_inode(comp->path.dentry);
> > +	int i;
> > +
> > +	list_for_each_entry(pos, &tpe->probes, list) {
> > +		orig = container_of(pos, struct trace_uprobe, tp);
> > +		if (comp_inode != d_real_inode(orig->path.dentry) ||
> > +		    comp->offset != orig->offset)
> > +			continue;
> > +
> > +		/*
> > +		 * trace_probe_compare_arg_type() ensured that nr_args and
> > +		 * each argument name and type are same. Let's compare comm.
> > +		 */
> > +		for (i = 0; i < orig->tp.nr_args; i++) {
> > +			if (strcmp(orig->tp.args[i].comm,
> > +				   comp->tp.args[i].comm))
> > +				continue;
> 
> Same as above.
> 
> > +		}
> > +
> > +		return true;
> > +	}
> > +
> > +	return false;
> > +}
> > +
> 
> -- 
> Thanks and Regards
> Srikar Dronamraju
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>