Re: [PATCH 1/6] ring-buffer: Prevent resizing of persistent ring buffer

[Vincent Donnefort <vdonnefort@google.com>]

On Mon, Aug 10, 2026 at 05:31:37PM +0900, Masami Hiramatsu wrote:
> On Fri, 7 Aug 2026 15:26:41 -0400
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > On Fri, 7 Aug 2026 16:45:23 +0100
> > Vincent Donnefort <vdonnefort@google.com> wrote:
> > 
> > > free_reserved_page() would do actually. But then it is definitive. 
> > 
> > It's not always a reserved page.
> > 
> > > 
> > > Happy to implement something like that. That also means that the instance can be
> > > actually freed?
> > 
> > They can be freed now. Try a rmdir on one.

I meant free_buffer_page() skips the memory for the persistent buffer.

> > 
> > Note implementing this is not straight forward. What I would suggest is

Yeah, I started looking at it and quickly understood :)

> > that because the persistent ring buffers are contiguous, to resize, you
> > basically need to remap to the new size. That means each of the buffers
> > will still be attached to each other.
> > 
> > What would need to be done is:
> > 
> >   1. calculate the new size needed to accommodate all the CPU buffers.
> >   2. Split them up within the new size region.
> >   3. Then free the remaining pages.
> > 
> > Obviously, access to the buffer from readers and writers will need to be
> > prohibited while this is happening.

That would mean losing all the data on the buffer? Actually the trace_remote
needs to teardown the whole buffer before the size is modified. Perhaps if
something similar is done for the persistent buffer, trace_remote could benefit.

> 
> Hmm, I think we also need to record the size of persistent ring buffer at
> initialization. The buffer size (number of pages are calculated by the
> size of reserved memory, which is defined in the kernel cmdline.

Could we store in ring_buffer_cpu_meta::buffers that the page is gone and must
be skipped to re-create the ring-buffer?

> 
> So if we have `reserve_mem=12M:4096:trace trace_instance=boot_map@trace` on
> the kernel cmdline, and we write the buffer_size = 1024(KB) on 2CPU machine,
> it can be shrinked down to ~2MB on the reserved memory. However, when we
> reboot the machine, the kernel calculates the size as 12MB again, and may
> get a validation failure.
> 
> So we also need to add nr_pages and nr_cpus (maybe) on ring_buffer_meta and
> calculate the ring buffer size from it, instead of using the reserved memory
> size (we also need to check the calculated size is smaller than that.)
> 
> Thank you,
> 
> -- 
> Masami Hiramatsu (Google) <mhiramat@kernel.org>

-- 
Vincent