Merge patch series "file: FD_{ADD,PREPARE}()"

Christian Brauner <brauner@kernel.org> says:

This now removes roughly double the code that it adds.

I've been playing with this to allow for moderately flexible usage of
the get_unused_fd_flags() + create file + fd_install() pattern that's
used quite extensively and requires cumbersome cleanup paths.

How callers allocate files is really heterogenous so it's not really
convenient to fold them into a single class. It's possibe to split them
into subclasses like for anon inodes. I think that's not necessarily
nice as well. This adds two primitives:

(1) FD_ADD() the simple cases a file is installed:

    fd = FD_ADD(O_CLOEXEC, vfio_device_open_file(device));
    if (fd < 0)
            vfio_device_put_registration(device);
    return fd;

(2) FD_PREPARE() that captures all the cases where access to fd or file
    or additional work before publishing the fd is needed:

    FD_PREPARE(fdf, O_CLOEXEC, sync_file->file);
    if (fdf.err) {
            fput(sync_file->file);
            return fdf.err;
    }

    data.fence = fd_prepare_fd(fdf);
    if (copy_to_user((void __user *)arg, &data, sizeof(data)))
            return -EFAULT;

    return fd_publish(fdf);

I've converted all of the easy cases over to it and it gets rid of an
aweful lot of convoluted cleanup logic. There are a bunch of other cases
that can also be converted after a bit of massaging.

It's centered around a simple struct. FD_PREPARE() encapsulates all of
allocation and cleanup logic and must be followed by a call to
fd_publish() which associates the fd with the file and installs it into
the callers fdtable. If fd_publish() isn't called both are deallocated.
FD_ADD() is a shorthand that does the fd_publish() and never exposes the
struct to the caller. That's often the case when they don't need access
to anything after installing the fd.

It mandates a specific order namely that first we allocate the fd and
then instantiate the file. But that shouldn't be a problem. Nearly
everyone I've converted used this order anyway.

There's a bunch of additional cases where it would be easy to convert
them to this pattern. For example, the whole sync file stuff in dma
currently returns the containing structure of the file instead of the
file itself even though it's only used to allocate files. Changing that
would make it fall into the FD_PREPARE() pattern easily. I've not done
that work yet.

There's room for extending this in a way that wed'd have subclasses for
some particularly often use patterns but as I said I'm not even sure
that's worth it.

* patches from https://patch.msgid.link/20251123-work-fd-prepare-v4-0-b6efa1706cfd@kernel.org: (47 commits)
  kvm: convert kvm_vcpu_ioctl_get_stats_fd() to FD_PREPARE()
  kvm: convert kvm_arch_supports_gmem_init_shared() to FD_PREPARE()
  io_uring: convert io_create_mock_file() to FD_PREPARE()
  file: convert replace_fd() to FD_PREPARE()
  vfio: convert vfio_group_ioctl_get_device_fd() to FD_PREPARE()
  tty: convert ptm_open_peer() to FD_PREPARE()
  ntsync: convert ntsync_obj_get_fd() to FD_PREPARE()
  media: convert media_request_alloc() to FD_PREPARE()
  hv: convert mshv_ioctl_create_partition() to FD_PREPARE()
  gpio: convert linehandle_create() to FD_PREPARE()
  dma: port sw_sync_ioctl_create_fence() to FD_PREPARE()
  pseries: port papr_rtas_setup_file_interface() to FD_PREPARE()
  pseries: convert papr_platform_dump_create_handle() to FD_PREPARE()
  spufs: convert spufs_gang_open() to FD_PREPARE()
  papr-hvpipe: convert papr_hvpipe_dev_create_handle() to FD_PREPARE()
  spufs: convert spufs_context_open() to FD_PREPARE()
  net/socket: convert __sys_accept4_file() to FD_PREPARE()
  net/socket: convert sock_map_fd() to FD_PREPARE()
  net/sctp: convert sctp_getsockopt_peeloff_common() to FD_PREPARE()
  net/kcm: convert kcm_ioctl() to FD_PREPARE()
  ...

Link: https://patch.msgid.link/20251123-work-fd-prepare-v4-0-b6efa1706cfd@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

15 files changed, 173 insertions, 366 deletions

diff --git a/fs/anon_inodes.c b/fs/anon_inodes.c
index 180a458fc4f7..b8381c7fb636 100644
--- a/fs/anon_inodes.c
+++ b/fs/anon_inodes.c
@@ -280,27 +280,8 @@ static int __anon_inode_getfd(const char *name,
 			      const struct inode *context_inode,
 			      bool make_inode)
 {
-	int error, fd;
-	struct file *file;
-
-	error = get_unused_fd_flags(flags);
-	if (error < 0)
-		return error;
-	fd = error;
-
-	file = __anon_inode_getfile(name, fops, priv, flags, context_inode,
-				    make_inode);
-	if (IS_ERR(file)) {
-		error = PTR_ERR(file);
-		goto err_put_unused_fd;
-	}
-	fd_install(fd, file);
-
-	return fd;
-
-err_put_unused_fd:
-	put_unused_fd(fd);
-	return error;
+	return FD_ADD(flags, __anon_inode_getfile(name, fops, priv, flags,
+						  context_inode, make_inode));
 }
 
 /**
diff --git a/fs/autofs/dev-ioctl.c b/fs/autofs/dev-ioctl.c
index d8dd150cbd74..ff67cd776812 100644
--- a/fs/autofs/dev-ioctl.c
+++ b/fs/autofs/dev-ioctl.c
@@ -231,32 +231,14 @@ static int test_by_type(const struct path *path, void *p)
  */
 static int autofs_dev_ioctl_open_mountpoint(const char *name, dev_t devid)
 {
-	int err, fd;
-
-	fd = get_unused_fd_flags(O_CLOEXEC);
-	if (likely(fd >= 0)) {
-		struct file *filp;
-		struct path path;
-
-		err = find_autofs_mount(name, &path, test_by_dev, &devid);
-		if (err)
-			goto out;
-
-		filp = dentry_open(&path, O_RDONLY, current_cred());
-		path_put(&path);
-		if (IS_ERR(filp)) {
-			err = PTR_ERR(filp);
-			goto out;
-		}
-
-		fd_install(fd, filp);
-	}
+	struct path path __free(path_put) = {};
+	int err;
 
-	return fd;
+	err = find_autofs_mount(name, &path, test_by_dev, &devid);
+	if (err)
+		return err;
 
-out:
-	put_unused_fd(fd);
-	return err;
+	return FD_ADD(O_CLOEXEC, dentry_open(&path, O_RDONLY, current_cred()));
 }
 
 /* Open a file descriptor on an autofs mount point */
diff --git a/fs/eventfd.c b/fs/eventfd.c
index af42b2c7d235..3219e0d596fe 100644
--- a/fs/eventfd.c
+++ b/fs/eventfd.c
@@ -378,9 +378,7 @@ EXPORT_SYMBOL_GPL(eventfd_ctx_fileget);
 
 static int do_eventfd(unsigned int count, int flags)
 {
-	struct eventfd_ctx *ctx;
-	struct file *file;
-	int fd;
+	struct eventfd_ctx *ctx __free(kfree) = NULL;
 
 	/* Check the EFD_* constants for consistency.  */
 	BUILD_BUG_ON(EFD_CLOEXEC != O_CLOEXEC);
@@ -398,26 +396,19 @@ static int do_eventfd(unsigned int count, int flags)
 	init_waitqueue_head(&ctx->wqh);
 	ctx->count = count;
 	ctx->flags = flags;
-	ctx->id = ida_alloc(&eventfd_ida, GFP_KERNEL);
 
 	flags &= EFD_SHARED_FCNTL_FLAGS;
 	flags |= O_RDWR;
-	fd = get_unused_fd_flags(flags);
-	if (fd < 0)
-		goto err;
-
-	file = anon_inode_getfile_fmode("[eventfd]", &eventfd_fops,
-					ctx, flags, FMODE_NOWAIT);
-	if (IS_ERR(file)) {
-		put_unused_fd(fd);
-		fd = PTR_ERR(file);
-		goto err;
-	}
-	fd_install(fd, file);
-	return fd;
-err:
-	eventfd_free_ctx(ctx);
-	return fd;
+
+	FD_PREPARE(fdf, flags,
+		   anon_inode_getfile_fmode("[eventfd]", &eventfd_fops, ctx,
+					    flags, FMODE_NOWAIT));
+	if (fdf.err)
+		return fdf.err;
+
+	ctx->id = ida_alloc(&eventfd_ida, GFP_KERNEL);
+	retain_and_null_ptr(ctx);
+	return fd_publish(fdf);
 }
 
 SYSCALL_DEFINE2(eventfd2, unsigned int, count, int, flags)
diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index ee7c4b683ec3..6c36d9dc6926 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -2165,9 +2165,8 @@ static void clear_tfile_check_list(void)
  */
 static int do_epoll_create(int flags)
 {
-	int error, fd;
-	struct eventpoll *ep = NULL;
-	struct file *file;
+	int error;
+	struct eventpoll *ep;
 
 	/* Check the EPOLL_* constant for consistency.  */
 	BUILD_BUG_ON(EPOLL_CLOEXEC != O_CLOEXEC);
@@ -2184,26 +2183,15 @@ static int do_epoll_create(int flags)
 	 * Creates all the items needed to setup an eventpoll file. That is,
 	 * a file structure and a free file descriptor.
 	 */
-	fd = get_unused_fd_flags(O_RDWR | (flags & O_CLOEXEC));
-	if (fd < 0) {
-		error = fd;
-		goto out_free_ep;
-	}
-	file = anon_inode_getfile("[eventpoll]", &eventpoll_fops, ep,
-				 O_RDWR | (flags & O_CLOEXEC));
-	if (IS_ERR(file)) {
-		error = PTR_ERR(file);
-		goto out_free_fd;
+	FD_PREPARE(fdf, O_RDWR | (flags & O_CLOEXEC),
+		   anon_inode_getfile("[eventpoll]", &eventpoll_fops, ep,
+				      O_RDWR | (flags & O_CLOEXEC)));
+	if (fdf.err) {
+		ep_clear_and_put(ep);
+		return fdf.err;
 	}
-	ep->file = file;
-	fd_install(fd, file);
-	return fd;
-
-out_free_fd:
-	put_unused_fd(fd);
-out_free_ep:
-	ep_clear_and_put(ep);
-	return error;
+	ep->file = fd_prepare_file(fdf);
+	return fd_publish(fdf);
 }
 
 SYSCALL_DEFINE1(epoll_create1, int, flags)
diff --git a/fs/exec.c b/fs/exec.c
index 6b70c6726d31..621898a26515 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1280,10 +1280,9 @@ int begin_new_exec(struct linux_binprm * bprm)
 
 	/* Pass the opened binary to the interpreter. */
 	if (bprm->have_execfd) {
-		retval = get_unused_fd_flags(0);
+		retval = FD_ADD(0, bprm->executable);
 		if (retval < 0)
 			goto out_unlock;
-		fd_install(retval, bprm->executable);
 		bprm->executable = NULL;
 		bprm->execfd = retval;
 	}
diff --git a/fs/fhandle.c b/fs/fhandle.c
index 052f9c9368fb..3de1547ec9d4 100644
--- a/fs/fhandle.c
+++ b/fs/fhandle.c
@@ -404,32 +404,28 @@ out_path:
 	return retval;
 }
 
+static struct file *file_open_handle(struct path *path, int open_flag)
+{
+	const struct export_operations *eops;
+
+	eops = path->mnt->mnt_sb->s_export_op;
+	if (eops->open)
+		return eops->open(path, open_flag);
+
+	return file_open_root(path, "", open_flag, 0);
+}
+
 static long do_handle_open(int mountdirfd, struct file_handle __user *ufh,
 			   int open_flag)
 {
-	long retval = 0;
+	long retval;
 	struct path path __free(path_put) = {};
-	struct file *file;
-	const struct export_operations *eops;
 
 	retval = handle_to_path(mountdirfd, ufh, &path, open_flag);
 	if (retval)
 		return retval;
 
-	CLASS(get_unused_fd, fd)(open_flag);
-	if (fd < 0)
-		return fd;
-
-	eops = path.mnt->mnt_sb->s_export_op;
-	if (eops->open)
-		file = eops->open(&path, open_flag);
-	else
-		file = file_open_root(&path, "", open_flag, 0);
-	if (IS_ERR(file))
-		return PTR_ERR(file);
-
-	fd_install(fd, file);
-	return take_fd(fd);
+	return FD_ADD(open_flag, file_open_handle(&path, open_flag));
 }
 
 /**
diff --git a/fs/file.c b/fs/file.c
index 28743b742e3c..7ea33a617896 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -1357,28 +1357,25 @@ out_unlock:
  */
 int receive_fd(struct file *file, int __user *ufd, unsigned int o_flags)
 {
-	int new_fd;
 	int error;
 
 	error = security_file_receive(file);
 	if (error)
 		return error;
 
-	new_fd = get_unused_fd_flags(o_flags);
-	if (new_fd < 0)
-		return new_fd;
+	FD_PREPARE(fdf, o_flags, file);
+	if (fdf.err)
+		return fdf.err;
+	get_file(file);
 
 	if (ufd) {
-		error = put_user(new_fd, ufd);
-		if (error) {
-			put_unused_fd(new_fd);
+		error = put_user(fd_prepare_fd(fdf), ufd);
+		if (error)
 			return error;
-		}
 	}
 
-	fd_install(new_fd, get_file(file));
-	__receive_sock(file);
-	return new_fd;
+	__receive_sock(fd_prepare_file(fdf));
+	return fd_publish(fdf);
 }
 EXPORT_SYMBOL_GPL(receive_fd);
 
diff --git a/fs/namespace.c b/fs/namespace.c
index d82910f33dc4..8302dd64be20 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -3103,19 +3103,7 @@ static struct file *vfs_open_tree(int dfd, const char __user *filename, unsigned
 
 SYSCALL_DEFINE3(open_tree, int, dfd, const char __user *, filename, unsigned, flags)
 {
-	int fd;
-	struct file *file __free(fput) = NULL;
-
-	file = vfs_open_tree(dfd, filename, flags);
-	if (IS_ERR(file))
-		return PTR_ERR(file);
-
-	fd = get_unused_fd_flags(flags & O_CLOEXEC);
-	if (fd < 0)
-		return fd;
-
-	fd_install(fd, no_free_ptr(file));
-	return fd;
+	return FD_ADD(flags, vfs_open_tree(dfd, filename, flags));
 }
 
 /*
@@ -4283,10 +4271,10 @@ static unsigned int attr_flags_to_mnt_flags(u64 attr_flags)
 SYSCALL_DEFINE3(fsmount, int, fs_fd, unsigned int, flags,
 		unsigned int, attr_flags)
 {
+	struct path new_path __free(path_put) = {};
 	struct mnt_namespace *ns;
 	struct fs_context *fc;
-	struct file *file;
-	struct path newmount;
+	struct vfsmount *new_mnt;
 	struct mount *mnt;
 	unsigned int mnt_flags = 0;
 	long ret;
@@ -4324,35 +4312,36 @@ SYSCALL_DEFINE3(fsmount, int, fs_fd, unsigned int, flags,
 
 	fc = fd_file(f)->private_data;
 
-	ret = mutex_lock_interruptible(&fc->uapi_mutex);
-	if (ret < 0)
+	ACQUIRE(mutex_intr, uapi_mutex)(&fc->uapi_mutex);
+	ret = ACQUIRE_ERR(mutex_intr, &uapi_mutex);
+	if (ret)
 		return ret;
 
 	/* There must be a valid superblock or we can't mount it */
 	ret = -EINVAL;
 	if (!fc->root)
-		goto err_unlock;
+		return ret;
 
 	ret = -EPERM;
 	if (mount_too_revealing(fc->root->d_sb, &mnt_flags)) {
 		errorfcp(fc, "VFS", "Mount too revealing");
-		goto err_unlock;
+		return ret;
 	}
 
 	ret = -EBUSY;
 	if (fc->phase != FS_CONTEXT_AWAITING_MOUNT)
-		goto err_unlock;
+		return ret;
 
 	if (fc->sb_flags & SB_MANDLOCK)
 		warn_mandlock();
 
-	newmount.mnt = vfs_create_mount(fc);
-	if (IS_ERR(newmount.mnt)) {
-		ret = PTR_ERR(newmount.mnt);
-		goto err_unlock;
-	}
-	newmount.dentry = dget(fc->root);
-	newmount.mnt->mnt_flags = mnt_flags;
+	new_mnt = vfs_create_mount(fc);
+	if (IS_ERR(new_mnt))
+		return PTR_ERR(new_mnt);
+	new_mnt->mnt_flags = mnt_flags;
+
+	new_path.dentry = dget(fc->root);
+	new_path.mnt = new_mnt;
 
 	/* We've done the mount bit - now move the file context into more or
 	 * less the same state as if we'd done an fspick().  We don't want to
@@ -4362,38 +4351,27 @@ SYSCALL_DEFINE3(fsmount, int, fs_fd, unsigned int, flags,
 	vfs_clean_context(fc);
 
 	ns = alloc_mnt_ns(current->nsproxy->mnt_ns->user_ns, true);
-	if (IS_ERR(ns)) {
-		ret = PTR_ERR(ns);
-		goto err_path;
-	}
-	mnt = real_mount(newmount.mnt);
+	if (IS_ERR(ns))
+		return PTR_ERR(ns);
+	mnt = real_mount(new_path.mnt);
 	ns->root = mnt;
 	ns->nr_mounts = 1;
 	mnt_add_to_ns(ns, mnt);
-	mntget(newmount.mnt);
+	mntget(new_path.mnt);
 
-	/* Attach to an apparent O_PATH fd with a note that we need to unmount
-	 * it, not just simply put it.
-	 */
-	file = dentry_open(&newmount, O_PATH, fc->cred);
-	if (IS_ERR(file)) {
-		dissolve_on_fput(newmount.mnt);
-		ret = PTR_ERR(file);
-		goto err_path;
+	FD_PREPARE(fdf, (flags & FSMOUNT_CLOEXEC) ? O_CLOEXEC : 0,
+		   dentry_open(&new_path, O_PATH, fc->cred));
+	if (fdf.err) {
+		dissolve_on_fput(new_path.mnt);
+		return fdf.err;
 	}
-	file->f_mode |= FMODE_NEED_UNMOUNT;
 
-	ret = get_unused_fd_flags((flags & FSMOUNT_CLOEXEC) ? O_CLOEXEC : 0);
-	if (ret >= 0)
-		fd_install(ret, file);
-	else
-		fput(file);
-
-err_path:
-	path_put(&newmount);
-err_unlock:
-	mutex_unlock(&fc->uapi_mutex);
-	return ret;
+	/*
+	 * Attach to an apparent O_PATH fd with a note that we
+	 * need to unmount it, not just simply put it.
+	 */
+	fd_prepare_file(fdf)->f_mode |= FMODE_NEED_UNMOUNT;
+	return fd_publish(fdf);
 }
 
 static inline int vfs_move_mount(const struct path *from_path,
@@ -5035,19 +5013,17 @@ SYSCALL_DEFINE5(open_tree_attr, int, dfd, const char __user *, filename,
 		unsigned, flags, struct mount_attr __user *, uattr,
 		size_t, usize)
 {
-	struct file __free(fput) *file = NULL;
-	int fd;
-
 	if (!uattr && usize)
 		return -EINVAL;
 
-	file = vfs_open_tree(dfd, filename, flags);
-	if (IS_ERR(file))
-		return PTR_ERR(file);
+	FD_PREPARE(fdf, flags, vfs_open_tree(dfd, filename, flags));
+	if (fdf.err)
+		return fdf.err;
 
 	if (uattr) {
-		int ret;
 		struct mount_kattr kattr = {};
+		struct file *file = fd_prepare_file(fdf);
+		int ret;
 
 		if (flags & OPEN_TREE_CLONE)
 			kattr.kflags = MOUNT_KATTR_IDMAP_REPLACE;
@@ -5063,12 +5039,7 @@ SYSCALL_DEFINE5(open_tree_attr, int, dfd, const char __user *, filename,
 			return ret;
 	}
 
-	fd = get_unused_fd_flags(flags & O_CLOEXEC);
-	if (fd < 0)
-		return fd;
-
-	fd_install(fd, no_free_ptr(file));
-	return fd;
+	return fd_publish(fdf);
 }
 
 int show_path(struct seq_file *m, struct dentry *root)
diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 1dadda82cae5..d0b9b984002f 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -1597,16 +1597,20 @@ static struct hlist_head *fanotify_alloc_merge_hash(void)
 	return hash;
 }
 
+DEFINE_CLASS(fsnotify_group,
+	     struct fsnotify_group *,
+	     if (!IS_ERR_OR_NULL(_T)) fsnotify_destroy_group(_T),
+	     fsnotify_alloc_group(ops, flags),
+	     const struct fsnotify_ops *ops, int flags)
+
 /* fanotify syscalls */
 SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 {
 	struct user_namespace *user_ns = current_user_ns();
-	struct fsnotify_group *group;
 	int f_flags, fd;
 	unsigned int fid_mode = flags & FANOTIFY_FID_BITS;
 	unsigned int class = flags & FANOTIFY_CLASS_BITS;
 	unsigned int internal_flags = 0;
-	struct file *file;
 
 	pr_debug("%s: flags=%x event_f_flags=%x\n",
 		 __func__, flags, event_f_flags);
@@ -1690,36 +1694,29 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 	if (flags & FAN_NONBLOCK)
 		f_flags |= O_NONBLOCK;
 
-	/* fsnotify_alloc_group takes a ref.  Dropped in fanotify_release */
-	group = fsnotify_alloc_group(&fanotify_fsnotify_ops,
+	CLASS(fsnotify_group, group)(&fanotify_fsnotify_ops,
 				     FSNOTIFY_GROUP_USER);
-	if (IS_ERR(group)) {
+	/* fsnotify_alloc_group takes a ref.  Dropped in fanotify_release */
+	if (IS_ERR(group))
 		return PTR_ERR(group);
-	}
 
 	/* Enforce groups limits per user in all containing user ns */
 	group->fanotify_data.ucounts = inc_ucount(user_ns, current_euid(),
 						  UCOUNT_FANOTIFY_GROUPS);
-	if (!group->fanotify_data.ucounts) {
-		fd = -EMFILE;
-		goto out_destroy_group;
-	}
+	if (!group->fanotify_data.ucounts)
+		return -EMFILE;
 
 	group->fanotify_data.flags = flags | internal_flags;
 	group->memcg = get_mem_cgroup_from_mm(current->mm);
 	group->user_ns = get_user_ns(user_ns);
 
 	group->fanotify_data.merge_hash = fanotify_alloc_merge_hash();
-	if (!group->fanotify_data.merge_hash) {
-		fd = -ENOMEM;
-		goto out_destroy_group;
-	}
+	if (!group->fanotify_data.merge_hash)
+		return -ENOMEM;
 
 	group->overflow_event = fanotify_alloc_overflow_event();
-	if (unlikely(!group->overflow_event)) {
-		fd = -ENOMEM;
-		goto out_destroy_group;
-	}
+	if (unlikely(!group->overflow_event))
+		return -ENOMEM;
 
 	if (force_o_largefile())
 		event_f_flags |= O_LARGEFILE;
@@ -1738,8 +1735,7 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 		group->priority = FSNOTIFY_PRIO_PRE_CONTENT;
 		break;
 	default:
-		fd = -EINVAL;
-		goto out_destroy_group;
+		return -EINVAL;
 	}
 
 	BUILD_BUG_ON(!(FANOTIFY_ADMIN_INIT_FLAGS & FAN_UNLIMITED_QUEUE));
@@ -1750,27 +1746,15 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 	}
 
 	if (flags & FAN_ENABLE_AUDIT) {
-		fd = -EPERM;
 		if (!capable(CAP_AUDIT_WRITE))
-			goto out_destroy_group;
-	}
-
-	fd = get_unused_fd_flags(f_flags);
-	if (fd < 0)
-		goto out_destroy_group;
-
-	file = anon_inode_getfile_fmode("[fanotify]", &fanotify_fops, group,
-					f_flags, FMODE_NONOTIFY);
-	if (IS_ERR(file)) {
-		put_unused_fd(fd);
-		fd = PTR_ERR(file);
-		goto out_destroy_group;
+			return -EPERM;
 	}
-	fd_install(fd, file);
-	return fd;
 
-out_destroy_group:
-	fsnotify_destroy_group(group);
+	fd = FD_ADD(f_flags,
+		    anon_inode_getfile_fmode("[fanotify]", &fanotify_fops,
+					     group, f_flags, FMODE_NONOTIFY));
+	if (fd >= 0)
+		retain_and_null_ptr(group);
 	return fd;
 }
 
diff --git a/fs/nsfs.c b/fs/nsfs.c
index 648dc59bef7f..375390bf810d 100644
--- a/fs/nsfs.c
+++ b/fs/nsfs.c
@@ -108,7 +108,6 @@ int ns_get_path(struct path *path, struct task_struct *task,
 int open_namespace(struct ns_common *ns)
 {
 	struct path path __free(path_put) = {};
-	struct file *f;
 	int err;
 
 	/* call first to consume reference */
@@ -116,16 +115,7 @@ int open_namespace(struct ns_common *ns)
 	if (err < 0)
 		return err;
 
-	CLASS(get_unused_fd, fd)(O_CLOEXEC);
-	if (fd < 0)
-		return fd;
-
-	f = dentry_open(&path, O_RDONLY, current_cred());
-	if (IS_ERR(f))
-		return PTR_ERR(f);
-
-	fd_install(fd, f);
-	return take_fd(fd);
+	return FD_ADD(O_CLOEXEC, dentry_open(&path, O_RDONLY, current_cred()));
 }
 
 int open_related_ns(struct ns_common *ns,
@@ -311,7 +301,6 @@ static long ns_ioctl(struct file *filp, unsigned int ioctl,
 		struct mnt_ns_info kinfo = {};
 		struct mnt_ns_info __user *uinfo = (struct mnt_ns_info __user *)arg;
 		struct path path __free(path_put) = {};
-		struct file *f __free(fput) = NULL;
 		size_t usize = _IOC_SIZE(ioctl);
 
 		if (ns->ns_type != CLONE_NEWNS)
@@ -330,28 +319,18 @@ static long ns_ioctl(struct file *filp, unsigned int ioctl,
 		if (ret)
 			return ret;
 
-		CLASS(get_unused_fd, fd)(O_CLOEXEC);
-		if (fd < 0)
-			return fd;
-
-		f = dentry_open(&path, O_RDONLY, current_cred());
-		if (IS_ERR(f))
-			return PTR_ERR(f);
-
-		if (uinfo) {
-			/*
-			 * If @uinfo is passed return all information about the
-			 * mount namespace as well.
-			 */
-			ret = copy_ns_info_to_user(to_mnt_ns(ns), uinfo, usize, &kinfo);
-			if (ret)
-				return ret;
-		}
-
-		/* Transfer reference of @f to caller's fdtable. */
-		fd_install(fd, no_free_ptr(f));
-		/* File descriptor is live so hand it off to the caller. */
-		return take_fd(fd);
+		FD_PREPARE(fdf, O_CLOEXEC, dentry_open(&path, O_RDONLY, current_cred()));
+		if (fdf.err)
+			return fdf.err;
+		/*
+		 * If @uinfo is passed return all information about the
+		 * mount namespace as well.
+		 */
+		ret = copy_ns_info_to_user(to_mnt_ns(ns), uinfo, usize, &kinfo);
+		if (ret)
+			return ret;
+		ret = fd_publish(fdf);
+		break;
 	}
 	default:
 		ret = -ENOTTY;
diff --git a/fs/open.c b/fs/open.c
index 3d64372ecc67..0ccd086323fd 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -1421,8 +1421,8 @@ static int do_sys_openat2(int dfd, const char __user *filename,
 			  struct open_how *how)
 {
 	struct open_flags op;
-	struct filename *tmp;
-	int err, fd;
+	struct filename *tmp __free(putname) = NULL;
+	int err;
 
 	err = build_open_flags(how, &op);
 	if (unlikely(err))
@@ -1432,18 +1432,7 @@ static int do_sys_openat2(int dfd, const char __user *filename,
 	if (IS_ERR(tmp))
 		return PTR_ERR(tmp);
 
-	fd = get_unused_fd_flags(how->flags);
-	if (likely(fd >= 0)) {
-		struct file *f = do_filp_open(dfd, tmp, &op);
-		if (IS_ERR(f)) {
-			put_unused_fd(fd);
-			fd = PTR_ERR(f);
-		} else {
-			fd_install(fd, f);
-		}
-	}
-	putname(tmp);
-	return fd;
+	return FD_ADD(how->flags, do_filp_open(dfd, tmp, &op));
 }
 
 int do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
diff --git a/fs/signalfd.c b/fs/signalfd.c
index d469782f97f4..d69eab584bc6 100644
--- a/fs/signalfd.c
+++ b/fs/signalfd.c
@@ -250,8 +250,6 @@ static const struct file_operations signalfd_fops = {
 
 static int do_signalfd4(int ufd, sigset_t *mask, int flags)
 {
-	struct signalfd_ctx *ctx;
-
 	/* Check the SFD_* constants for consistency.  */
 	BUILD_BUG_ON(SFD_CLOEXEC != O_CLOEXEC);
 	BUILD_BUG_ON(SFD_NONBLOCK != O_NONBLOCK);
@@ -263,7 +261,8 @@ static int do_signalfd4(int ufd, sigset_t *mask, int flags)
 	signotset(mask);
 
 	if (ufd == -1) {
-		struct file *file;
+		int fd;
+		struct signalfd_ctx *ctx __free(kfree) = NULL;
 
 		ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
 		if (!ctx)
@@ -271,22 +270,16 @@ static int do_signalfd4(int ufd, sigset_t *mask, int flags)
 
 		ctx->sigmask = *mask;
 
-		ufd = get_unused_fd_flags(flags & O_CLOEXEC);
-		if (ufd < 0) {
-			kfree(ctx);
-			return ufd;
-		}
-
-		file = anon_inode_getfile_fmode("[signalfd]", &signalfd_fops,
-					ctx, O_RDWR | (flags & O_NONBLOCK),
-					FMODE_NOWAIT);
-		if (IS_ERR(file)) {
-			put_unused_fd(ufd);
-			kfree(ctx);
-			return PTR_ERR(file);
-		}
-		fd_install(ufd, file);
+		fd = FD_ADD(flags & O_CLOEXEC,
+			    anon_inode_getfile_fmode(
+				    "[signalfd]", &signalfd_fops, ctx,
+				    O_RDWR | (flags & O_NONBLOCK), FMODE_NOWAIT));
+		if (fd >= 0)
+			retain_and_null_ptr(ctx);
+		return fd;
 	} else {
+		struct signalfd_ctx *ctx;
+
 		CLASS(fd, f)(ufd);
 		if (fd_empty(f))
 			return -EBADF;
diff --git a/fs/timerfd.c b/fs/timerfd.c
index c68f28d9c426..9fcea7860ddf 100644
--- a/fs/timerfd.c
+++ b/fs/timerfd.c
@@ -393,9 +393,8 @@ static const struct file_operations timerfd_fops = {
 
 SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
 {
-	int ufd;
-	struct timerfd_ctx *ctx;
-	struct file *file;
+	struct timerfd_ctx *ctx __free(kfree) = NULL;
+	int ret;
 
 	/* Check the TFD_* constants for consistency.  */
 	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
@@ -432,23 +431,13 @@ SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
 
 	ctx->moffs = ktime_mono_to_real(0);
 
-	ufd = get_unused_fd_flags(flags & TFD_SHARED_FCNTL_FLAGS);
-	if (ufd < 0) {
-		kfree(ctx);
-		return ufd;
-	}
-
-	file = anon_inode_getfile_fmode("[timerfd]", &timerfd_fops, ctx,
-			    O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS),
-			    FMODE_NOWAIT);
-	if (IS_ERR(file)) {
-		put_unused_fd(ufd);
-		kfree(ctx);
-		return PTR_ERR(file);
-	}
-
-	fd_install(ufd, file);
-	return ufd;
+	ret = FD_ADD(flags & TFD_SHARED_FCNTL_FLAGS,
+		     anon_inode_getfile_fmode("[timerfd]", &timerfd_fops, ctx,
+					      O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS),
+					      FMODE_NOWAIT));
+	if (ret >= 0)
+		retain_and_null_ptr(ctx);
+	return ret;
 }
 
 static int do_timerfd_settime(int ufd, int flags, 
diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index 54c6cc7fe9c6..e6e74b384087 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -2111,9 +2111,7 @@ static void init_once_userfaultfd_ctx(void *mem)
 
 static int new_userfaultfd(int flags)
 {
-	struct userfaultfd_ctx *ctx;
-	struct file *file;
-	int fd;
+	struct userfaultfd_ctx *ctx __free(kfree) = NULL;
 
 	VM_WARN_ON_ONCE(!current->mm);
 
@@ -2135,26 +2133,18 @@ static int new_userfaultfd(int flags)
 	atomic_set(&ctx->mmap_changing, 0);
 	ctx->mm = current->mm;
 
-	fd = get_unused_fd_flags(flags & UFFD_SHARED_FCNTL_FLAGS);
-	if (fd < 0)
-		goto err_out;
+	FD_PREPARE(fdf, flags & UFFD_SHARED_FCNTL_FLAGS,
+		   anon_inode_create_getfile("[userfaultfd]", &userfaultfd_fops, ctx,
+					     O_RDONLY | (flags & UFFD_SHARED_FCNTL_FLAGS),
+					     NULL));
+	if (fdf.err)
+		return fdf.err;
 
-	/* Create a new inode so that the LSM can block the creation.  */
-	file = anon_inode_create_getfile("[userfaultfd]", &userfaultfd_fops, ctx,
-			O_RDONLY | (flags & UFFD_SHARED_FCNTL_FLAGS), NULL);
-	if (IS_ERR(file)) {
-		put_unused_fd(fd);
-		fd = PTR_ERR(file);
-		goto err_out;
-	}
 	/* prevent the mm struct to be freed */
 	mmgrab(ctx->mm);
-	file->f_mode |= FMODE_NOWAIT;
-	fd_install(fd, file);
-	return fd;
-err_out:
-	kmem_cache_free(userfaultfd_ctx_cachep, ctx);
-	return fd;
+	fd_prepare_file(fdf)->f_mode |= FMODE_NOWAIT;
+	retain_and_null_ptr(ctx);
+	return fd_publish(fdf);
 }
 
 static inline bool userfaultfd_syscall_allowed(int flags)
diff --git a/fs/xfs/xfs_handle.c b/fs/xfs/xfs_handle.c
index f19fce557354..5a3e3bf4e7cc 100644
--- a/fs/xfs/xfs_handle.c
+++ b/fs/xfs/xfs_handle.c
@@ -233,14 +233,11 @@ xfs_open_by_handle(
 	xfs_fsop_handlereq_t	*hreq)
 {
 	const struct cred	*cred = current_cred();
-	int			error;
-	int			fd;
 	int			permflag;
-	struct file		*filp;
 	struct inode		*inode;
 	struct dentry		*dentry;
 	fmode_t			fmode;
-	struct path		path;
+	struct path		path __free(path_put) = {};
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
@@ -249,12 +246,11 @@ xfs_open_by_handle(
 	if (IS_ERR(dentry))
 		return PTR_ERR(dentry);
 	inode = d_inode(dentry);
+	path.dentry = dentry;
 
 	/* Restrict xfs_open_by_handle to directories & regular files. */
-	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
-		error = -EPERM;
-		goto out_dput;
-	}
+	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
+		return -EPERM;
 
 #if BITS_PER_LONG != 32
 	hreq->oflags |= O_LARGEFILE;
@@ -263,48 +259,30 @@ xfs_open_by_handle(
 	permflag = hreq->oflags;
 	fmode = OPEN_FMODE(permflag);
 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
-	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
-		error = -EPERM;
-		goto out_dput;
-	}
+	    (fmode & FMODE_WRITE) && IS_APPEND(inode))
+		return -EPERM;
 
-	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
-		error = -EPERM;
-		goto out_dput;
-	}
+	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode))
+		return -EPERM;
 
 	/* Can't write directories. */
-	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
-		error = -EISDIR;
-		goto out_dput;
-	}
+	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE))
+		return -EISDIR;
 
-	fd = get_unused_fd_flags(0);
-	if (fd < 0) {
-		error = fd;
-		goto out_dput;
-	}
+	path.mnt = mntget(parfilp->f_path.mnt);
 
-	path.mnt = parfilp->f_path.mnt;
-	path.dentry = dentry;
-	filp = dentry_open(&path, hreq->oflags, cred);
-	dput(dentry);
-	if (IS_ERR(filp)) {
-		put_unused_fd(fd);
-		return PTR_ERR(filp);
-	}
+	FD_PREPARE(fdf, 0, dentry_open(&path, hreq->oflags, cred));
+	if (fdf.err)
+		return fdf.err;
 
 	if (S_ISREG(inode->i_mode)) {
+		struct file *filp = fd_prepare_file(fdf);
+
 		filp->f_flags |= O_NOATIME;
 		filp->f_mode |= FMODE_NOCMTIME;
 	}
 
-	fd_install(fd, filp);
-	return fd;
-
- out_dput:
-	dput(dentry);
-	return error;
+	return fd_publish(fdf);
 }
 
 int