On Tue, Sep 09, 2025 at 02:52:36PM -0400, Steven Rostedt wrote:
> On Tue, 9 Sep 2025 14:39:48 -0400
> Steven Rostedt <[email protected]> wrote:
> 
> > But anyway, I think it should work for the remote buffers too. Let me go
> > and fix the current iterator.
> 
> I thought it was broken but it isn't ;-) I did it properly, I just didn't
> look deep enough.
> 
> So yeah, look at the rb_iter_head_event() code. The ring buffer iterator
> has a copy of the event. It has:
> 
>       if ((iter->head + length) > commit || length > iter->event_size)
>               /* Writer corrupted the read? */
>               goto reset;
> 
>       memcpy(iter->event, event, length);
>       /*
>        * If the page stamp is still the same after this rmb() then the
>        * event was safely copied without the writer entering the page.
>        */
>       smp_rmb();
> 
>       /* Make sure the page didn't change since we read this */
>       if (iter->page_stamp != iter_head_page->page->time_stamp ||
>           commit > rb_page_commit(iter_head_page))
>               goto reset;
> 
> It first checks before copying that the data it's about to copy hasn't been
> touched by the writer.
> 
> It then copies the event into the iter->event temp buffer.
> 
> Then it checks again that the data hasn't been touched. If it has, then
> consider the data corrupt and end the iteration. This is how the "trace"
> file works. I believe you could do the same for the remote code.
> 
> If we are gonna keep the "trace" file, let's make sure it's fully
> implemented.

I was more worry about the ring-buffer page order that can be reshuffled on each
swap_reader_page(), making the page links useless in the kernel. Ideally, the
meta-page would keep the page ID order somewhere.

Alternatively, we could walk all the buffer pages to read the timestamp and
re-create the order but that sounds quite cumbersome.

> 
> -- Steve

Reply via email to