Re: [PATCH 3/4] ring-buffer: Drop redundant ring_buffer_iter::event_size

[Steven Rostedt <rostedt@goodmis.org>]

On Tue, 22 Sep 2026 08:29:43 +0100
Vincent Donnefort <vdonnefort@google.com> wrote:

> event_size contains the maximum size the iterator can read for each
> event. It corresponds to the maximum capacity of a ring-buffer page.
> During non-consuming read, resizing is disabled. It is therefore safe to
> just use rb_subbuf_capacity(), making event_size redundant. Remove it.

Hmm, I don't like the code dependency here. Yes, we disable resizing and
the event_size isn't going to change, but if in the future that is no
longer the case, then this becomes a hidden bug.

What's wrong with just having a field that keeps track of the size
allocated? It's more robust that way and doesn't have any hidden
dependencies.

-- Steve
 

> 
> Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
> ---
>  kernel/trace/ring_buffer.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
> index 1a158a24b3c8..3129589e2d0e 100644
> --- a/kernel/trace/ring_buffer.c
> +++ b/kernel/trace/ring_buffer.c
> @@ -680,7 +680,6 @@ struct ring_buffer_iter {
>  	u64				read_stamp;
>  	u64				page_stamp;
>  	struct ring_buffer_event	*event;
> -	size_t				event_size;
>  	int				missed_events;
>  };
>  
> @@ -3558,9 +3557,11 @@ rb_iter_head_event(struct ring_buffer_iter *iter)
>  	 */
>  	barrier();
>  
> -	if ((iter->head + length) > commit || length > iter->event_size)
> +	if ((iter->head + length) > commit ||
> +	    length > rb_subbuf_capacity(iter->cpu_buffer->buffer)) {
>  		/* Writer corrupted the read? */
>  		goto reset;
> +	}
>  
>  	memcpy(iter->event, event, length);
>  	/*
> @@ -6488,8 +6489,7 @@ ring_buffer_read_start(struct trace_buffer *buffer, int cpu, gfp_t flags)
>  		mutex_unlock(&buffer->mutex);
>  
>  	/* Holds the entire event: data and meta data. */
> -	iter->event_size = rb_page_capacity(READ_ONCE(cpu_buffer->reader_page));
> -	iter->event = kmalloc(iter->event_size, flags);
> +	iter->event = kmalloc(rb_subbuf_capacity(buffer), flags);
>  	if (!iter->event) {
>  		atomic_dec(&cpu_buffer->resize_disabled);
>  		return NULL;