Re: [PATCH] tracing: fprobe: fix suspicious rcu usage in fprobe_entry

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

On Thu, 04 Sep 2025 11:37:35 +0800
Menglong Dong <menglong.dong@linux.dev> wrote:

> On 2025/9/3 12:22 Herbert Xu <herbert@gondor.apana.org.au> write:
> > On Tue, Sep 02, 2025 at 05:50:32PM +0800, menglong.dong@linux.dev wrote:
> > > On 2025/9/2 17:17 Herbert Xu <herbert@gondor.apana.org.au> write:
> > > > Menglong Dong <dongml2@chinatelecom.cn> wrote:
> > > > >
> > > > > diff --git a/kernel/trace/fprobe.c b/kernel/trace/fprobe.c
> > > > > index fb127fa95f21..fece0f849c1c 100644
> > > > > --- a/kernel/trace/fprobe.c
> > > > > +++ b/kernel/trace/fprobe.c
> > > > > @@ -269,7 +269,9 @@ static int fprobe_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
> > > > >        if (WARN_ON_ONCE(!fregs))
> > > > >                return 0;
> > > > > 
> > > > > +       rcu_read_lock();
> > > > >        head = rhltable_lookup(&fprobe_ip_table, &func, fprobe_rht_params);
> > > > > +       rcu_read_unlock();
> > > > >        reserved_words = 0;
> > > > >        rhl_for_each_entry_rcu(node, pos, head, hlist) {
> > > > >                if (node->addr != func)
> > > > 
> > > > Actually this isn't quite right.  I know that it is a false-positive
> > > > so that it's actually safe, but if you're going to mark it with
> > > > rcu_read_lock, it should cover both the lookup as well as the
> > > > dereference which happens in the loop rhl_for_each_entry_rcu.
> > > 
> > > Yeah, I understand. The rcu_read_lock() here is totally used to
> > > suppress the suspicious rcu usage warning, not for the protection.
> > > So I used it just for the rhltable_lookup() to reduce the impact.
> > > Maybe I should add some comment for it.
> > 
> > My point is that after a lookup you will be doing some sort of a
> > dereference on the RCU pointer.  That would cause exactly the same
> > splat that rhltable_lookup itself generated.
> > 
> > For example, rhl_for_each_entry_rcu should have created the same
> > warning, but it doesn't because for some reason it is using
> > rcu_dereference_raw.  I'll need to dig up the history of this
> > to see if there is a good reason for it to not warn.
> 
> Yeah, I understand what you mean. I noticed this, and that's why
> I added the rcu_read_lock() for rhashtable_lookup() only.
> 
> Maybe it is to obtain better performance? Just guess ;)
> And hlist_for_each_entry_rcu() also uses rcu_dereference_raw().

Hi Menglong, if you update the patch to use guard(rcu)() because
head is used repeatedly in fprobe_entry(), I can replace it.

Thank you,

> 
> Thanks!
> Menglong Dong
> > 
> > Cheers,
> > -- 
> > Email: Herbert Xu <herbert@gondor.apana.org.au>
> > Home Page: http://gondor.apana.org.au/~herbert/
> > PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
> > 
> 
> 
> 
> 


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