On Tue, Sep 06, 2022 at 07:58:50AM +0300, Amir Goldstein wrote: > On Mon, Sep 5, 2022 at 7:25 PM Miklos Szeredi <miklos@xxxxxxxxxx> wrote: > > > > On Wed, Aug 31, 2022 at 02:30:40PM -0700, Yu-li Lin wrote: > > > Thanks for the reference. IIUC, the consensus is to make it atomic, > > > although there's no agreement on how it should be done. Does that mean > > > we should hold off on > > > this patch until atomic temp files are figured out higher in the stack > > > or do you have thoughts on how the fuse uapi should look like prior to > > > the vfs/refactoring decision? > > > > Here's a patch refactoring the tmpfile kapi to return an open file instead of a > > dentry. > > > > Comments? > > IDGI. Why did you need to place do_dentry_open() in all the implementations > and not inside vfs_tmpfile_new()? > Am I missing something? The whole point of that horror is to have open done inside ->tmpfile() instances... Al, very unhappy with proposed interface ;-/
