On Fri, Dec 13, 2019 at 7:36 PM Jens Axboe <axboe@xxxxxxxxx> wrote: > Just one caller of this, and just use filp_close() there manually. > This is important to allow async close/removal of the fd. [...] > index 3da91a112bab..a250d291c71b 100644 > --- a/fs/file.c > +++ b/fs/file.c > @@ -662,7 +662,7 @@ int __close_fd_get_file(unsigned int fd, struct file **res) > spin_unlock(&files->file_lock); > get_file(file); > *res = file; > - return filp_close(file, files); > + return 0; Above this function is a comment saying "variant of __close_fd that gets a ref on the file for later fput"; that should probably be changed to point out that you also still need to filp_close().
