On Thu, Feb 15, 2024 at 8:27 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote: > I feel there's a lot more you could say here. Like "There are N files > which only include mm.h in order to use folio_next()". You can see this detail in the last patch (14/14), but if you prefer to have this spelled out here, I'll amend the patch description. Though I only worked on the headers, not the thousands of sources which might also be able to reduce their header dependencies. Getting the exact number of those is a huge effort, and it's worth it, but not today. Once this patch series is merged, we can incrementally start this, but these patches are the precondition. > Otherwise ... > why is this more than churn? This is about code correctness and faster compile times. You may or may not value that, much of this is really just opinion. Similar efforts have been merged recently, which is why I thought this goal was consensus among kernel people.
