Re: [PATCH v14 4/5] ring-buffer: Reset RB_MISSED_* flags on persistent ring buffer

[Steven Rostedt <rostedt@goodmis.org>]

On Mon, 30 Mar 2026 21:50:20 +0900
"Masami Hiramatsu (Google)" <mhiramat@kernel.org> wrote:

> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> 
> Reset RB_MISSED_* flags when the persistent ring buffer is
> validated at boot. Since these flags are used only in reading
> process, such process should be stopped when reboot and never
> be restarted. Thus, these flags are meaningless in the next
> boot. Moreover, it can confuse the read process after reboot.

Is it meaningless on a second boot?

Let's say you have a crash, and there's an invalid buffer. On the next boot
it is flagged as invalid with the RB_MISSED flag. But then you reboot again
before looking at the buffer. The next boot will clear this flag. Now
looking at the persistent ring buffer will not show any missed events.

Ideally, it shouldn't matter how many reboots are made. If the persistent
ring buffer hasn't started again, it should always show the same output.

-- Steve


> 
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> ---
>  Changes in v14:
>    - Newly added.
> ---
>  kernel/trace/ring_buffer.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
> index e5178239f2f9..5049cf13021e 100644
> --- a/kernel/trace/ring_buffer.c
> +++ b/kernel/trace/ring_buffer.c
> @@ -1903,6 +1903,7 @@ static int rb_validate_buffer(struct buffer_page *bpage, int cpu,
>  		local_set(&bpage->page->commit, 0);
>  	} else {
>  		local_set(&bpage->entries, ret);
> +		local_set(&bpage->page->commit, tail);
>  	}
>  
>  	return ret;