On Wed, 10 Sep 2025 10:45:05 +0100 Vincent Donnefort <vdonnef...@google.com> wrote:
> > 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. Yeah, internally we could keep the order of the pages. Shouldn't be too hard, as the swapping happens by the kernel. > > Alternatively, we could walk all the buffer pages to read the timestamp and > re-create the order but that sounds quite cumbersome. No, having a separate array of the order is probably what we want. Then every swap_reader_page() will update the array. The trace iterator could simply walk that array. -- Steve
