struct fd: representation change

	We want the compiler to see that fdput() on empty instance
is a no-op.  The emptiness check is that file reference is NULL,
while fdput() is "fput() if FDPUT_FPUT is present in flags".
The reason why fdput() on empty instance is a no-op is something
compiler can't see - it's that we never generate instances with
NULL file reference combined with non-zero flags.

	It's not that hard to deal with - the real primitives behind
fdget() et.al. are returning an unsigned long value, unpacked by (inlined)
__to_fd() into the current struct file * + int.  The lower bits are
used to store flags, while the rest encodes the pointer.  Linus suggested
that keeping this unsigned long around with the extractions done by inlined
accessors should generate a sane code and that turns out to be the case.
Namely, turning struct fd into a struct-wrapped unsinged long, with
        fd_empty(f) => unlikely(f.word == 0)
	fd_file(f) => (struct file *)(f.word & ~3)
	fdput(f) => if (f.word & 1) fput(fd_file(f))
ends up with compiler doing the right thing.  The cost is the patch
footprint, of course - we need to switch f.file to fd_file(f) all over
the tree, and it's not doable with simple search and replace; there are
false positives, etc.

	Note that the sole member of that structure is an opaque
unsigned long - all accesses should be done via wrappers and I don't
want to use a name that would invite manual casts to file pointers,
etc.  The value of that member is equal either to (unsigned long)p | flags,
p being an address of some struct file instance, or to 0 for an empty fd.

	For now the new predicate (fd_empty(f)) has no users; all the
existing checks have form (!fd_file(f)).  We will convert to fd_empty()
use later; here we only define it (and tell the compiler that it's
unlikely to return true).

	This commit only deals with representation change; there will
be followups.

Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Author: Al Viro

drivers/infiniband/core/uverbs_cmd.c

@@ -572,7 +572,7 @@ static int ib_uverbs_open_xrcd(struct uverbs_attr_bundle *attrs)
 	struct inode                   *inode = NULL;
 	int				new_xrcd = 0;
 	struct ib_device *ib_dev;
-	struct fd f = {};
+	struct fd f = EMPTY_FD;
 	int ret;
 
 	ret = uverbs_request(attrs, &cmd, sizeof(cmd));

fs/overlayfs/file.c

@@ -93,11 +93,11 @@ static int ovl_real_fdget_meta(const struct file *file, struct fd *real,
 			       bool allow_meta)
 {
 	struct dentry *dentry = file_dentry(file);
+	struct file *realfile = file->private_data;
 	struct path realpath;
 	int err;
 
-	real->flags = 0;
-	real->file = file->private_data;
+	real->word = (unsigned long)realfile;
 
 	if (allow_meta) {
 		ovl_path_real(dentry, &realpath);
@@ -113,27 +113,29 @@ static int ovl_real_fdget_meta(const struct file *file, struct fd *real,
 		return -EIO;
 
 	/* Has it been copied up since we'd opened it? */
-	if (unlikely(file_inode(real->file) != d_inode(realpath.dentry))) {
-		real->flags = FDPUT_FPUT;
-		real->file = ovl_open_realfile(file, &realpath);
-
-		return PTR_ERR_OR_ZERO(real->file);
+	if (unlikely(file_inode(realfile) != d_inode(realpath.dentry))) {
+		struct file *f = ovl_open_realfile(file, &realpath);
+		if (IS_ERR(f))
+			return PTR_ERR(f);
+		real->word = (unsigned long)ovl_open_realfile(file, &realpath) | FDPUT_FPUT;
+		return 0;
 	}
 
 	/* Did the flags change since open? */
-	if (unlikely((file->f_flags ^ real->file->f_flags) & ~OVL_OPEN_FLAGS))
-		return ovl_change_flags(real->file, file->f_flags);
+	if (unlikely((file->f_flags ^ realfile->f_flags) & ~OVL_OPEN_FLAGS))
+		return ovl_change_flags(realfile, file->f_flags);
 
 	return 0;
 }
 
 static int ovl_real_fdget(const struct file *file, struct fd *real)
 {
 	if (d_is_dir(file_dentry(file))) {
-		real->flags = 0;
-		real->file = ovl_dir_real_file(file, false);
-
-		return PTR_ERR_OR_ZERO(real->file);
+		struct file *f = ovl_dir_real_file(file, false);
+		if (IS_ERR(f))
+			return PTR_ERR(f);
+		real->word = (unsigned long)f;
+		return 0;
 	}
 
 	return ovl_real_fdget_meta(file, real, false);

fs/xfs/xfs_handle.c

@@ -85,7 +85,7 @@ xfs_find_handle(
 	int			hsize;
 	xfs_handle_t		handle;
 	struct inode		*inode;
-	struct fd		f = {NULL};
+	struct fd		f = EMPTY_FD;
 	struct path		path;
 	int			error;
 	struct xfs_inode	*ip;

include/linux/file.h

@@ -35,18 +35,28 @@ static inline void fput_light(struct file *file, int fput_needed)
 		fput(file);
 }
 
+/* either a reference to struct file + flags
+ * (cloned vs. borrowed, pos locked), with
+ * flags stored in lower bits of value,
+ * or empty (represented by 0).
+ */
 struct fd {
-	struct file *file;
-	unsigned int flags;
+	unsigned long word;
 };
 #define FDPUT_FPUT       1
 #define FDPUT_POS_UNLOCK 2
 
-#define fd_file(f) ((f).file)
+#define fd_file(f) ((struct file *)((f).word & ~(FDPUT_FPUT|FDPUT_POS_UNLOCK)))
+static inline bool fd_empty(struct fd f)
+{
+	return unlikely(!f.word);
+}
+
+#define EMPTY_FD (struct fd){0}
 
 static inline void fdput(struct fd fd)
 {
-	if (fd.flags & FDPUT_FPUT)
+	if (fd.word & FDPUT_FPUT)
 		fput(fd_file(fd));
 }
 
@@ -60,7 +70,7 @@ extern void __f_unlock_pos(struct file *);
 
 static inline struct fd __to_fd(unsigned long v)
 {
-	return (struct fd){(struct file *)(v & ~3),v & 3};
+	return (struct fd){v};
 }
 
 static inline struct fd fdget(unsigned int fd)
@@ -80,7 +90,7 @@ static inline struct fd fdget_pos(int fd)
 
 static inline void fdput_pos(struct fd f)
 {
-	if (f.flags & FDPUT_POS_UNLOCK)
+	if (f.word & FDPUT_POS_UNLOCK)
 		__f_unlock_pos(fd_file(f));
 	fdput(f);
 }

kernel/events/core.c

@@ -12474,7 +12474,7 @@ SYSCALL_DEFINE5(perf_event_open,
 	struct perf_event_attr attr;
 	struct perf_event_context *ctx;
 	struct file *event_file = NULL;
-	struct fd group = {NULL, 0};
+	struct fd group = EMPTY_FD;
 	struct task_struct *task = NULL;
 	struct pmu *pmu;
 	int event_fd;

net/socket.c

@@ -559,7 +559,7 @@ static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 	if (fd_file(f)) {
 		sock = sock_from_file(fd_file(f));
 		if (likely(sock)) {
-			*fput_needed = f.flags & FDPUT_FPUT;
+			*fput_needed = f.word & FDPUT_FPUT;
 			return sock;
 		}
 		*err = -ENOTSOCK;