[PATCH v18 3/8] ring-buffer: Skip invalid sub-buffers when validating persistent ring buffer

[Masami Hiramatsu (Google)]

On Tue, 28 Apr 2026 15:55:08 -0400
Steven Rostedt <rostedt at goodmis.org> wrote:

> On Fri, 24 Apr 2026 15:52:27 +0900
> "Masami Hiramatsu (Google)" <mhiramat at kernel.org> wrote:
> 
> > @@ -5648,11 +5668,12 @@ __rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
> >   again:
> >  	/*
> >  	 * This should normally only loop twice. But because the
> > -	 * start of the reader inserts an empty page, it causes
> > -	 * a case where we will loop three times. There should be no
> > -	 * reason to loop four times (that I know of).
> > +	 * start of the reader inserts an empty page, it causes a
> > +	 * case where we will loop three times. There should be no
> > +	 * reason to loop four times unless the ring buffer is a
> > +	 * recovered persistent ring buffer.
> 
> Can you explain more to why this is allowed for persistent ring buffer?

Ah, that was introduced in v15.

  Changes in v15:
  - Skip reader_page loop check on persistent ring buffer because
    there can be contiguous empty(invalidated) pages. 


So, finding next reader_page, we need to skip empty pages, which is normally
not contiguous. However, if we see more than 3 invalid pages on recovering
persistent ring buffer, it will be reset and become empty.


> 
> Note, I do not like any loops that can go into an infinite loop and lock up
> the machine. If something goes wrong with a persistent ring buffer, then
> this could possibly go into an infinite loop.

Yeah, so I think we should not use goto here. OK, let me update it to
an actual loop.

> 
> I want to understand why this is allowed, and possibly add a check that
> prevents this from never ending.

It should not be a never ending loop (there are other exit conditions),
but I agreed. What about limiting with nr_subbufs?

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 5326924615a4..aa89ec96e964 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -5630,7 +5630,8 @@ __rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 	 * a case where we will loop three times. There should be no
 	 * reason to loop four times (that I know of).
 	 */
-	if (RB_WARN_ON(cpu_buffer, ++nr_loops > 3)) {
+	if (RB_WARN_ON(cpu_buffer,
+		++nr_loops > (cpu_buffer->ring_meta ? cpu_buffer->nr_subbufs : 3))) {
 		reader = NULL;
 		goto out;
 	}

> 
> -- Steve
> 
> 
> >  	 */
> > -	if (RB_WARN_ON(cpu_buffer, ++nr_loops > 3)) {
> > +	if (RB_WARN_ON(cpu_buffer, ++nr_loops > 3 && !cpu_buffer->ring_meta)) {
> >  		reader = NULL;
> >  		goto out;
> >  	}
> 


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