On Wed, May 08, 2024 at 10:14:44AM -0700, Linus Torvalds wrote: > On Wed, 8 May 2024 at 09:19, Linus Torvalds > <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > So since we already have two versions of F_DUPFD (the other being > > F_DUPFD_CLOEXEC) I decided that the best thing to do is to just extend > > on that existing naming pattern, and called it F_DUPFD_QUERY instead. > > > > I'm not married to the name, so if somebody hates it, feel free to > > argue otherwise. > > Side note: with this patch, doing > > ret = fcntl(fd1, F_DUPFD_QUERY, fd2); > > will result in: > > -1 (EBADF): 'fd1' is not a valid file descriptor > -1 (EINVAL): old kernel that doesn't support F_DUPFD_QUERY > 0: fd2 does not refer to the same file as fd1 > 1: fd2 is the same 'struct file' as fd1 > > and it might be worth noting a couple of things here: > > (a) fd2 being an invalid file descriptor does not cause EBADF, it > just causes "does not match". > > (b) we *could* use more bits for more equality > > IOW, it would possibly make sense to extend the 0/1 result to be > > - bit #0: same file pointer > - bit #1: same path > - bit #2: same dentry > - bit #3: same inode > > which are all different levels of "sameness". Not worth it without someone explaining in detail why imho. First pass should be to try and replace kcmp() in scenarios where it's obviously not needed or overkill. I've added a CLASS(fd_raw) in a preliminary patch since we'll need that anyway which means that your comparison patch becomes even simpler imho. I've also added a selftest patch: https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git/log/?h=vfs.misc ?
