Re: [PATCH] tracing: fprobe: use ftrace if CONFIG_DYNAMIC_FTRACE_WITH_ARGS

[Menglong Dong <menglong.dong@linux.dev>]

On 2025/10/30 08:40, Masami Hiramatsu wrote:
> On Wed, 29 Oct 2025 10:15:14 +0800
> Menglong Dong <menglong8.dong@gmail.com> wrote:
> 
> > For now, we will use ftrace for the fprobe if fp->exit_handler not exists
> > and CONFIG_DYNAMIC_FTRACE_WITH_REGS is enabled.
> > 
> > However, CONFIG_DYNAMIC_FTRACE_WITH_REGS is not supported by some arch,
> > such as arm. What we need in the fprobe is the function arguments, so we
> > can use ftrace for fprobe if CONFIG_DYNAMIC_FTRACE_WITH_ARGS is enabled.
> > 
> > Therefore, use ftrace if CONFIG_DYNAMIC_FTRACE_WITH_REGS or
> > CONFIG_DYNAMIC_FTRACE_WITH_ARGS enabled.
> > 
> > Signed-off-by: Menglong Dong <dongml2@chinatelecom.cn>
> > ---
> >  kernel/trace/fprobe.c | 12 ++++++++----
> >  1 file changed, 8 insertions(+), 4 deletions(-)
> > 
> > diff --git a/kernel/trace/fprobe.c b/kernel/trace/fprobe.c
> > index ecd623eef68b..9fad0569f521 100644
> > --- a/kernel/trace/fprobe.c
> > +++ b/kernel/trace/fprobe.c
> > @@ -254,7 +254,11 @@ static inline int __fprobe_kprobe_handler(unsigned long ip, unsigned long parent
> >  	return ret;
> >  }
> >  
> > -#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> > +#if defined(CONFIG_DYNAMIC_FTRACE_WITH_ARGS) || defined(CONFIG_DYNAMIC_FTRACE_WITH_REGS)
> > +#define FPROBE_USE_FTRACE
> > +#endif
> > +
> > +#ifdef FPROBE_USE_FTRACE
> >  /* ftrace_ops callback, this processes fprobes which have only entry_handler. */
> >  static void fprobe_ftrace_entry(unsigned long ip, unsigned long parent_ip,
> >  	struct ftrace_ops *ops, struct ftrace_regs *fregs)
> > @@ -295,7 +299,7 @@ NOKPROBE_SYMBOL(fprobe_ftrace_entry);
> >  
> >  static struct ftrace_ops fprobe_ftrace_ops = {
> >  	.func	= fprobe_ftrace_entry,
> > -	.flags	= FTRACE_OPS_FL_SAVE_REGS,
> > +	.flags	= FTRACE_OPS_FL_SAVE_ARGS,
> >  };
> >  static int fprobe_ftrace_active;
> >  
> > @@ -349,7 +353,7 @@ static bool fprobe_is_ftrace(struct fprobe *fp)
> >  {
> >  	return false;
> >  }
> > -#endif
> > +#endif /* !FPROBE_USE_FTRACE */
> >  
> >  /* fgraph_ops callback, this processes fprobes which have exit_handler. */
> >  static int fprobe_fgraph_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
> > @@ -599,7 +603,7 @@ static int fprobe_module_callback(struct notifier_block *nb,
> >  	if (alist.index > 0) {
> >  		ftrace_set_filter_ips(&fprobe_graph_ops.ops,
> >  				      alist.addrs, alist.index, 1, 0);
> > -#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> > +#ifdef FPROBE_USE_FTRACE
> >  		ftrace_set_filter_ips(&fprobe_ftrace_ops,
> >  				      alist.addrs, alist.index, 1, 0);
> >  #endif
> 
> Instead of introducing FPROBE_USE_FTRACE, I think it is better to add another
> wrapper to be called here.

Sounds better! I'll modify it in the next version.

Thanks!
Menglong Dong

> 
> #if defined(CONFIG_DYNAMIC_FTRACE_WITH_ARGS) || defined(CONFIG_DYNAMIC_FTRACE_WITH_REGS)
> 
> static void fprobe_set_ips(unsigned long *ips, unsigned int cnt, int remove, int reset)
> {
> 		ftrace_set_filter_ips(&fprobe_graph_ops.ops,
> 				      alist.addrs, alist.index, 1, 0);
> 		ftrace_set_filter_ips(&fprobe_ftrace_ops,
> 				      alist.addrs, alist.index, 1, 0);
> }
> ...
> #else
> static void fprobe_set_ips(unsigned long *ips, unsigned int cnt, int remove, int reset)
> {
> 		ftrace_set_filter_ips(&fprobe_graph_ops.ops,
> 				      alist.addrs, alist.index, 1, 0);
> }
> #endif
> 
> Thank you,
> 
>