On Wed, Sep 16, 2020 at 06:21:26AM +0100, Christoph Hellwig wrote: > On Wed, Sep 16, 2020 at 04:27:17AM +0100, Matthew Wilcox (Oracle) wrote: > > The nr_thps counter is to support THPs in the page cache when the > > filesystem doesn't understand THPs. Eventually it will be removed, but > > we should still support filesystems which do not understand THPs yet. > > Move the nr_thp manipulation functions to filemap.h since they're > > page-cache specific. > > Honestly I don't think we should support the read-only THP crap. We > should in fact never have merged that bandaid to start with given that > you did good progress on the real thing. I'd like to see the feature ripped out again, yes. Once we have a few more filesystems converted, I think that'll be a reasonable thing to do. It was a good step along the way; Song fixed a number of problems, and worked on other things that I never had to learn anything about (like uprobes and khugepaged). I wouldn't go so far as to say we should never have merged it, but I think we can remove it in about six months.
