Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> writes: > On Tue, 8 Aug 2023 at 10:15, Christian Brauner <brauner@xxxxxxxxxx> wrote: >> >> I think you're at least missing the removal of the PF_KTHREAD check > > Yup. > >> It'd be neat to leave that in so >> __fput_sync() doesn't get proliferated to non PF_KTHREAD without us >> noticing. So maybe we just need a tiny primitive. > > Considering that over the decade we've had this, we've only grown two > cases of actually using it, I think we're fine. That and two cases of flush_delayed_fput() followed by task_work_run(). That combined with a maintainer who was actively against any new calls to __fput_sync and a version of __fput_sync that called BUG_ON if you used it. So I am not 100% convinced that there were so few calls to __fput_sync simply because people couldn't think of a need for it. Eric
