Merge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs

Pull overlayfs updates from Miklos Szeredi:
 "This work from Amir introduces the inodes index feature, which
  provides:

   - hardlinks are not broken on copy up

   - infrastructure for overlayfs NFS export

  This also fixes constant st_ino for samefs case for lower hardlinks"

* 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs: (33 commits)
  ovl: mark parent impure and restore timestamp on ovl_link_up()
  ovl: document copying layers restrictions with inodes index
  ovl: cleanup orphan index entries
  ovl: persistent overlay inode nlink for indexed inodes
  ovl: implement index dir copy up
  ovl: move copy up lock out
  ovl: rearrange copy up
  ovl: add flag for upper in ovl_entry
  ovl: use struct copy_up_ctx as function argument
  ovl: base tmpfile in workdir too
  ovl: factor out ovl_copy_up_inode() helper
  ovl: extract helper to get temp file in copy up
  ovl: defer upper dir lock to tempfile link
  ovl: hash overlay non-dir inodes by copy up origin
  ovl: cleanup bad and stale index entries on mount
  ovl: lookup index entry for copy up origin
  ovl: verify index dir matches upper dir
  ovl: verify upper root dir matches lower root dir
  ovl: introduce the inodes index dir feature
  ovl: generalize ovl_create_workdir()
  ...

Documentation/filesystems/overlayfs.txt

@@ -201,6 +201,40 @@ rightmost one and going left.  In the above example lower1 will be the
 top, lower2 the middle and lower3 the bottom layer.
 
 
+Sharing and copying layers
+--------------------------
+
+Lower layers may be shared among several overlay mounts and that is indeed
+a very common practice.  An overlay mount may use the same lower layer
+path as another overlay mount and it may use a lower layer path that is
+beneath or above the path of another overlay lower layer path.
+
+Using an upper layer path and/or a workdir path that are already used by
+another overlay mount is not allowed and will fail with EBUSY.  Using
+partially overlapping paths is not allowed but will not fail with EBUSY.
+
+Mounting an overlay using an upper layer path, where the upper layer path
+was previously used by another mounted overlay in combination with a
+different lower layer path, is allowed, unless the "inodes index" feature
+is enabled.
+
+With the "inodes index" feature, on the first time mount, an NFS file
+handle of the lower layer root directory, along with the UUID of the lower
+filesystem, are encoded and stored in the "trusted.overlay.origin" extended
+attribute on the upper layer root directory.  On subsequent mount attempts,
+the lower root directory file handle and lower filesystem UUID are compared
+to the stored origin in upper root directory.  On failure to verify the
+lower root origin, mount will fail with ESTALE.  An overlayfs mount with
+"inodes index" enabled will fail with EOPNOTSUPP if the lower filesystem
+does not support NFS export, lower filesystem does not have a valid UUID or
+if the upper filesystem does not support extended attributes.
+
+It is quite a common practice to copy overlay layers to a different
+directory tree on the same or different underlying filesystem, and even
+to a different machine.  With the "inodes index" feature, trying to mount
+the copied layers will fail the verification of the lower root file handle.
+
+
 Non-standard behavior
 ---------------------
 

fs/overlayfs/Kconfig

@@ -23,3 +23,23 @@ config OVERLAY_FS_REDIRECT_DIR
 	  Note, that redirects are not backward compatible.  That is, mounting
 	  an overlay which has redirects on a kernel that doesn't support this
 	  feature will have unexpected results.
+
+config OVERLAY_FS_INDEX
+	bool "Overlayfs: turn on inodes index feature by default"
+	depends on OVERLAY_FS
+	help
+	  If this config option is enabled then overlay filesystems will use
+	  the inodes index dir to map lower inodes to upper inodes by default.
+	  In this case it is still possible to turn off index globally with the
+	  "index=off" module option or on a filesystem instance basis with the
+	  "index=off" mount option.
+
+	  The inodes index feature prevents breaking of lower hardlinks on copy
+	  up.
+
+	  Note, that the inodes index feature is read-only backward compatible.
+	  That is, mounting an overlay which has an index dir on a kernel that
+	  doesn't support this feature read-only, will not have any negative
+	  outcomes.  However, mounting the same overlay with an old kernel
+	  read-write and then mounting it again with a new kernel, will have
+	  unexpected results.

fs/overlayfs/copy_up.c

@@ -233,12 +233,13 @@ int ovl_set_attr(struct dentry *upperdentry, struct kstat *stat)
 	return err;
 }
 
-static struct ovl_fh *ovl_encode_fh(struct dentry *lower, uuid_t *uuid)
+struct ovl_fh *ovl_encode_fh(struct dentry *lower, bool is_upper)
 {
 	struct ovl_fh *fh;
 	int fh_type, fh_len, dwords;
 	void *buf;
 	int buflen = MAX_HANDLE_SZ;
+	uuid_t *uuid = &lower->d_sb->s_uuid;
 
 	buf = kmalloc(buflen, GFP_TEMPORARY);
 	if (!buf)
@@ -271,6 +272,14 @@ static struct ovl_fh *ovl_encode_fh(struct dentry *lower, uuid_t *uuid)
 	fh->magic = OVL_FH_MAGIC;
 	fh->type = fh_type;
 	fh->flags = OVL_FH_FLAG_CPU_ENDIAN;
+	/*
+	 * When we will want to decode an overlay dentry from this handle
+	 * and all layers are on the same fs, if we get a disconncted real
+	 * dentry when we decode fid, the only way to tell if we should assign
+	 * it to upperdentry or to lowerstack is by checking this flag.
+	 */
+	if (is_upper)
+		fh->flags |= OVL_FH_FLAG_PATH_UPPER;
 	fh->len = fh_len;
 	fh->uuid = *uuid;
 	memcpy(fh->fid, buf, buflen);
@@ -283,7 +292,6 @@ out:
 static int ovl_set_origin(struct dentry *dentry, struct dentry *lower,
 			  struct dentry *upper)
 {
-	struct super_block *sb = lower->d_sb;
 	const struct ovl_fh *fh = NULL;
 	int err;
 
@@ -292,9 +300,8 @@ static int ovl_set_origin(struct dentry *dentry, struct dentry *lower,
 	 * so we can use the overlay.origin xattr to distignuish between a copy
 	 * up and a pure upper inode.
 	 */
-	if (sb->s_export_op && sb->s_export_op->fh_to_dentry &&
-	    !uuid_is_null(&sb->s_uuid)) {
-		fh = ovl_encode_fh(lower, &sb->s_uuid);
+	if (ovl_can_decode_fh(lower->d_sb)) {
+		fh = ovl_encode_fh(lower, false);
 		if (IS_ERR(fh))
 			return PTR_ERR(fh);
 	}
@@ -309,84 +316,156 @@ static int ovl_set_origin(struct dentry *dentry, struct dentry *lower,
 	return err;
 }
 
-static int ovl_copy_up_locked(struct dentry *workdir, struct dentry *upperdir,
-			      struct dentry *dentry, struct path *lowerpath,
-			      struct kstat *stat, const char *link,
-			      struct kstat *pstat, bool tmpfile)
+struct ovl_copy_up_ctx {
+	struct dentry *parent;
+	struct dentry *dentry;
+	struct path lowerpath;
+	struct kstat stat;
+	struct kstat pstat;
+	const char *link;
+	struct dentry *destdir;
+	struct qstr destname;
+	struct dentry *workdir;
+	bool tmpfile;
+	bool origin;
+};
+
+static int ovl_link_up(struct ovl_copy_up_ctx *c)
+{
+	int err;
+	struct dentry *upper;
+	struct dentry *upperdir = ovl_dentry_upper(c->parent);
+	struct inode *udir = d_inode(upperdir);
+
+	/* Mark parent "impure" because it may now contain non-pure upper */
+	err = ovl_set_impure(c->parent, upperdir);
+	if (err)
+		return err;
+
+	err = ovl_set_nlink_lower(c->dentry);
+	if (err)
+		return err;
+
+	inode_lock_nested(udir, I_MUTEX_PARENT);
+	upper = lookup_one_len(c->dentry->d_name.name, upperdir,
+			       c->dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (!IS_ERR(upper)) {
+		err = ovl_do_link(ovl_dentry_upper(c->dentry), udir, upper,
+				  true);
+		dput(upper);
+
+		if (!err) {
+			/* Restore timestamps on parent (best effort) */
+			ovl_set_timestamps(upperdir, &c->pstat);
+			ovl_dentry_set_upper_alias(c->dentry);
+		}
+	}
+	inode_unlock(udir);
+	ovl_set_nlink_upper(c->dentry);
+
+	return err;
+}
+
+static int ovl_install_temp(struct ovl_copy_up_ctx *c, struct dentry *temp,
+			    struct dentry **newdentry)
 {
-	struct inode *wdir = workdir->d_inode;
-	struct inode *udir = upperdir->d_inode;
-	struct dentry *newdentry = NULL;
-	struct dentry *upper = NULL;
-	struct dentry *temp = NULL;
 	int err;
+	struct dentry *upper;
+	struct inode *udir = d_inode(c->destdir);
+
+	upper = lookup_one_len(c->destname.name, c->destdir, c->destname.len);
+	if (IS_ERR(upper))
+		return PTR_ERR(upper);
+
+	if (c->tmpfile)
+		err = ovl_do_link(temp, udir, upper, true);
+	else
+		err = ovl_do_rename(d_inode(c->workdir), temp, udir, upper, 0);
+
+	if (!err)
+		*newdentry = dget(c->tmpfile ? upper : temp);
+	dput(upper);
+
+	return err;
+}
+
+static int ovl_get_tmpfile(struct ovl_copy_up_ctx *c, struct dentry **tempp)
+{
+	int err;
+	struct dentry *temp;
 	const struct cred *old_creds = NULL;
 	struct cred *new_creds = NULL;
 	struct cattr cattr = {
 		/* Can't properly set mode on creation because of the umask */
-		.mode = stat->mode & S_IFMT,
-		.rdev = stat->rdev,
-		.link = link
+		.mode = c->stat.mode & S_IFMT,
+		.rdev = c->stat.rdev,
+		.link = c->link
 	};
 
-	err = security_inode_copy_up(dentry, &new_creds);
+	err = security_inode_copy_up(c->dentry, &new_creds);
 	if (err < 0)
 		goto out;
 
 	if (new_creds)
 		old_creds = override_creds(new_creds);
 
-	if (tmpfile)
-		temp = ovl_do_tmpfile(upperdir, stat->mode);
-	else
-		temp = ovl_lookup_temp(workdir);
-	err = 0;
-	if (IS_ERR(temp)) {
-		err = PTR_ERR(temp);
-		temp = NULL;
+	if (c->tmpfile) {
+		temp = ovl_do_tmpfile(c->workdir, c->stat.mode);
+		if (IS_ERR(temp))
+			goto temp_err;
+	} else {
+		temp = ovl_lookup_temp(c->workdir);
+		if (IS_ERR(temp))
+			goto temp_err;
+
+		err = ovl_create_real(d_inode(c->workdir), temp, &cattr,
+				      NULL, true);
+		if (err) {
+			dput(temp);
+			goto out;
+		}
 	}
-
-	if (!err && !tmpfile)
-		err = ovl_create_real(wdir, temp, &cattr, NULL, true);
-
+	err = 0;
+	*tempp = temp;
+out:
 	if (new_creds) {
 		revert_creds(old_creds);
 		put_cred(new_creds);
 	}
 
-	if (err)
-		goto out;
+	return err;
 
-	if (S_ISREG(stat->mode)) {
+temp_err:
+	err = PTR_ERR(temp);
+	goto out;
+}
+
+static int ovl_copy_up_inode(struct ovl_copy_up_ctx *c, struct dentry *temp)
+{
+	int err;
+
+	if (S_ISREG(c->stat.mode)) {
 		struct path upperpath;
 
-		ovl_path_upper(dentry, &upperpath);
+		ovl_path_upper(c->dentry, &upperpath);
 		BUG_ON(upperpath.dentry != NULL);
 		upperpath.dentry = temp;
 
-		if (tmpfile) {
-			inode_unlock(udir);
-			err = ovl_copy_up_data(lowerpath, &upperpath,
-					       stat->size);
-			inode_lock_nested(udir, I_MUTEX_PARENT);
-		} else {
-			err = ovl_copy_up_data(lowerpath, &upperpath,
-					       stat->size);
-		}
-
+		err = ovl_copy_up_data(&c->lowerpath, &upperpath, c->stat.size);
 		if (err)
-			goto out_cleanup;
+			return err;
 	}
 
-	err = ovl_copy_xattr(lowerpath->dentry, temp);
+	err = ovl_copy_xattr(c->lowerpath.dentry, temp);
 	if (err)
-		goto out_cleanup;
+		return err;
 
 	inode_lock(temp->d_inode);
-	err = ovl_set_attr(temp, stat);
+	err = ovl_set_attr(temp, &c->stat);
 	inode_unlock(temp->d_inode);
 	if (err)
-		goto out_cleanup;
+		return err;
 
 	/*
 	 * Store identifier of lower inode in upper inode xattr to
@@ -395,41 +474,48 @@ static int ovl_copy_up_locked(struct dentry *workdir, struct dentry *upperdir,
 	 * Don't set origin when we are breaking the association with a lower
 	 * hard link.
 	 */
-	if (S_ISDIR(stat->mode) || stat->nlink == 1) {
-		err = ovl_set_origin(dentry, lowerpath->dentry, temp);
+	if (c->origin) {
+		err = ovl_set_origin(c->dentry, c->lowerpath.dentry, temp);
 		if (err)
-			goto out_cleanup;
+			return err;
 	}
 
-	upper = lookup_one_len(dentry->d_name.name, upperdir,
-			       dentry->d_name.len);
-	if (IS_ERR(upper)) {
-		err = PTR_ERR(upper);
-		upper = NULL;
-		goto out_cleanup;
-	}
+	return 0;
+}
 
-	if (tmpfile)
-		err = ovl_do_link(temp, udir, upper, true);
-	else
-		err = ovl_do_rename(wdir, temp, udir, upper, 0);
+static int ovl_copy_up_locked(struct ovl_copy_up_ctx *c)
+{
+	struct inode *udir = c->destdir->d_inode;
+	struct dentry *newdentry = NULL;
+	struct dentry *temp = NULL;
+	int err;
+
+	err = ovl_get_tmpfile(c, &temp);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up_inode(c, temp);
 	if (err)
 		goto out_cleanup;
 
-	newdentry = dget(tmpfile ? upper : temp);
-	ovl_dentry_update(dentry, newdentry);
-	ovl_inode_update(d_inode(dentry), d_inode(newdentry));
+	if (c->tmpfile) {
+		inode_lock_nested(udir, I_MUTEX_PARENT);
+		err = ovl_install_temp(c, temp, &newdentry);
+		inode_unlock(udir);
+	} else {
+		err = ovl_install_temp(c, temp, &newdentry);
+	}
+	if (err)
+		goto out_cleanup;
 
-	/* Restore timestamps on parent (best effort) */
-	ovl_set_timestamps(upperdir, pstat);
+	ovl_inode_update(d_inode(c->dentry), newdentry);
 out:
 	dput(temp);
-	dput(upper);
 	return err;
 
 out_cleanup:
-	if (!tmpfile)
-		ovl_cleanup(wdir, temp);
+	if (!c->tmpfile)
+		ovl_cleanup(d_inode(c->workdir), temp);
 	goto out;
 }
 
@@ -442,78 +528,119 @@ out_cleanup:
  * is possible that the copy up will lock the old parent.  At that point
  * the file will have already been copied up anyway.
  */
+static int ovl_do_copy_up(struct ovl_copy_up_ctx *c)
+{
+	int err;
+	struct ovl_fs *ofs = c->dentry->d_sb->s_fs_info;
+	bool indexed = false;
+
+	if (ovl_indexdir(c->dentry->d_sb) && !S_ISDIR(c->stat.mode) &&
+	    c->stat.nlink > 1)
+		indexed = true;
+
+	if (S_ISDIR(c->stat.mode) || c->stat.nlink == 1 || indexed)
+		c->origin = true;
+
+	if (indexed) {
+		c->destdir = ovl_indexdir(c->dentry->d_sb);
+		err = ovl_get_index_name(c->lowerpath.dentry, &c->destname);
+		if (err)
+			return err;
+	} else {
+		/*
+		 * Mark parent "impure" because it may now contain non-pure
+		 * upper
+		 */
+		err = ovl_set_impure(c->parent, c->destdir);
+		if (err)
+			return err;
+	}
+
+	/* Should we copyup with O_TMPFILE or with workdir? */
+	if (S_ISREG(c->stat.mode) && ofs->tmpfile) {
+		c->tmpfile = true;
+		err = ovl_copy_up_locked(c);
+	} else {
+		err = -EIO;
+		if (lock_rename(c->workdir, c->destdir) != NULL) {
+			pr_err("overlayfs: failed to lock workdir+upperdir\n");
+		} else {
+			err = ovl_copy_up_locked(c);
+			unlock_rename(c->workdir, c->destdir);
+		}
+	}
+
+	if (indexed) {
+		if (!err)
+			ovl_set_flag(OVL_INDEX, d_inode(c->dentry));
+		kfree(c->destname.name);
+	} else if (!err) {
+		struct inode *udir = d_inode(c->destdir);
+
+		/* Restore timestamps on parent (best effort) */
+		inode_lock(udir);
+		ovl_set_timestamps(c->destdir, &c->pstat);
+		inode_unlock(udir);
+
+		ovl_dentry_set_upper_alias(c->dentry);
+	}
+
+	return err;
+}
+
 static int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
-			   struct path *lowerpath, struct kstat *stat)
+			   int flags)
 {
-	DEFINE_DELAYED_CALL(done);
-	struct dentry *workdir = ovl_workdir(dentry);
 	int err;
-	struct kstat pstat;
+	DEFINE_DELAYED_CALL(done);
 	struct path parentpath;
-	struct dentry *lowerdentry = lowerpath->dentry;
-	struct dentry *upperdir;
-	const char *link = NULL;
-	struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
+	struct ovl_copy_up_ctx ctx = {
+		.parent = parent,
+		.dentry = dentry,
+		.workdir = ovl_workdir(dentry),
+	};
 
-	if (WARN_ON(!workdir))
+	if (WARN_ON(!ctx.workdir))
 		return -EROFS;
 
-	ovl_do_check_copy_up(lowerdentry);
-
-	ovl_path_upper(parent, &parentpath);
-	upperdir = parentpath.dentry;
-
-	/* Mark parent "impure" because it may now contain non-pure upper */
-	err = ovl_set_impure(parent, upperdir);
+	ovl_path_lower(dentry, &ctx.lowerpath);
+	err = vfs_getattr(&ctx.lowerpath, &ctx.stat,
+			  STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
 	if (err)
 		return err;
 
-	err = vfs_getattr(&parentpath, &pstat,
+	ovl_path_upper(parent, &parentpath);
+	ctx.destdir = parentpath.dentry;
+	ctx.destname = dentry->d_name;
+
+	err = vfs_getattr(&parentpath, &ctx.pstat,
 			  STATX_ATIME | STATX_MTIME, AT_STATX_SYNC_AS_STAT);
 	if (err)
 		return err;
 
-	if (S_ISLNK(stat->mode)) {
-		link = vfs_get_link(lowerdentry, &done);
-		if (IS_ERR(link))
-			return PTR_ERR(link);
-	}
-
-	/* Should we copyup with O_TMPFILE or with workdir? */
-	if (S_ISREG(stat->mode) && ofs->tmpfile) {
-		err = ovl_copy_up_start(dentry);
-		/* err < 0: interrupted, err > 0: raced with another copy-up */
-		if (unlikely(err)) {
-			pr_debug("ovl_copy_up_start(%pd2) = %i\n", dentry, err);
-			if (err > 0)
-				err = 0;
-			goto out_done;
-		}
-
-		inode_lock_nested(upperdir->d_inode, I_MUTEX_PARENT);
-		err = ovl_copy_up_locked(workdir, upperdir, dentry, lowerpath,
-					 stat, link, &pstat, true);
-		inode_unlock(upperdir->d_inode);
-		ovl_copy_up_end(dentry);
-		goto out_done;
-	}
+	/* maybe truncate regular file. this has no effect on dirs */
+	if (flags & O_TRUNC)
+		ctx.stat.size = 0;
 
-	err = -EIO;
-	if (lock_rename(workdir, upperdir) != NULL) {
-		pr_err("overlayfs: failed to lock workdir+upperdir\n");
-		goto out_unlock;
+	if (S_ISLNK(ctx.stat.mode)) {
+		ctx.link = vfs_get_link(ctx.lowerpath.dentry, &done);
+		if (IS_ERR(ctx.link))
+			return PTR_ERR(ctx.link);
 	}
-	if (ovl_dentry_upper(dentry)) {
-		/* Raced with another copy-up?  Nothing to do, then... */
-		err = 0;
-		goto out_unlock;
+	ovl_do_check_copy_up(ctx.lowerpath.dentry);
+
+	err = ovl_copy_up_start(dentry);
+	/* err < 0: interrupted, err > 0: raced with another copy-up */
+	if (unlikely(err)) {
+		if (err > 0)
+			err = 0;
+	} else {
+		if (!ovl_dentry_upper(dentry))
+			err = ovl_do_copy_up(&ctx);
+		if (!err && !ovl_dentry_has_upper_alias(dentry))
+			err = ovl_link_up(&ctx);
+		ovl_copy_up_end(dentry);
 	}
-
-	err = ovl_copy_up_locked(workdir, upperdir, dentry, lowerpath,
-				 stat, link, &pstat, false);
-out_unlock:
-	unlock_rename(workdir, upperdir);
-out_done:
 	do_delayed_call(&done);
 
 	return err;
@@ -527,11 +654,22 @@ int ovl_copy_up_flags(struct dentry *dentry, int flags)
 	while (!err) {
 		struct dentry *next;
 		struct dentry *parent;
-		struct path lowerpath;
-		struct kstat stat;
-		enum ovl_path_type type = ovl_path_type(dentry);
 
-		if (OVL_TYPE_UPPER(type))
+		/*
+		 * Check if copy-up has happened as well as for upper alias (in
+		 * case of hard links) is there.
+		 *
+		 * Both checks are lockless:
+		 *  - false negatives: will recheck under oi->lock
+		 *  - false positives:
+		 *    + ovl_dentry_upper() uses memory barriers to ensure the
+		 *      upper dentry is up-to-date
+		 *    + ovl_dentry_has_upper_alias() relies on locking of
+		 *      upper parent i_rwsem to prevent reordering copy-up
+		 *      with rename.
+		 */
+		if (ovl_dentry_upper(dentry) &&
+		    ovl_dentry_has_upper_alias(dentry))
 			break;
 
 		next = dget(dentry);
@@ -539,22 +677,14 @@ int ovl_copy_up_flags(struct dentry *dentry, int flags)
 		for (;;) {
 			parent = dget_parent(next);
 
-			type = ovl_path_type(parent);
-			if (OVL_TYPE_UPPER(type))
+			if (ovl_dentry_upper(parent))
 				break;
 
 			dput(next);
 			next = parent;
 		}
 
-		ovl_path_lower(next, &lowerpath);
-		err = vfs_getattr(&lowerpath, &stat,
-				  STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
-		/* maybe truncate regular file. this has no effect on dirs */
-		if (flags & O_TRUNC)
-			stat.size = 0;
-		if (!err)
-			err = ovl_copy_up_one(parent, next, &lowerpath, &stat);
+		err = ovl_copy_up_one(parent, next, flags);
 
 		dput(parent);
 		dput(next);

fs/overlayfs/dir.c

@@ -24,7 +24,7 @@ module_param_named(redirect_max, ovl_redirect_max, ushort, 0644);
 MODULE_PARM_DESC(ovl_redirect_max,
 		 "Maximum length of absolute redirect xattr value");
 
-void ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
+int ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
 {
 	int err;
 
@@ -39,6 +39,8 @@ void ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
 		pr_err("overlayfs: cleanup of '%pd2' failed (%i)\n",
 		       wdentry, err);
 	}
+
+	return err;
 }
 
 struct dentry *ovl_lookup_temp(struct dentry *workdir)
@@ -154,12 +156,13 @@ static void ovl_instantiate(struct dentry *dentry, struct inode *inode,
 			    struct dentry *newdentry, bool hardlink)
 {
 	ovl_dentry_version_inc(dentry->d_parent);
-	ovl_dentry_update(dentry, newdentry);
+	ovl_dentry_set_upper_alias(dentry);
 	if (!hardlink) {
-		ovl_inode_update(inode, d_inode(newdentry));
+		ovl_inode_update(inode, newdentry);
 		ovl_copyattr(newdentry->d_inode, inode);
 	} else {
-		WARN_ON(ovl_inode_real(inode, NULL) != d_inode(newdentry));
+		WARN_ON(ovl_inode_real(inode) != d_inode(newdentry));
+		dput(newdentry);
 		inc_nlink(inode);
 	}
 	d_instantiate(dentry, inode);
@@ -588,6 +591,7 @@ static int ovl_link(struct dentry *old, struct inode *newdir,
 		    struct dentry *new)
 {
 	int err;
+	bool locked = false;
 	struct inode *inode;
 
 	err = ovl_want_write(old);
@@ -598,6 +602,10 @@ static int ovl_link(struct dentry *old, struct inode *newdir,
 	if (err)
 		goto out_drop_write;
 
+	err = ovl_nlink_start(old, &locked);
+	if (err)
+		goto out_drop_write;
+
 	inode = d_inode(old);
 	ihold(inode);
 
@@ -605,12 +613,18 @@ static int ovl_link(struct dentry *old, struct inode *newdir,
 	if (err)
 		iput(inode);
 
+	ovl_nlink_end(old, locked);
 out_drop_write:
 	ovl_drop_write(old);
 out:
 	return err;
 }
 
+static bool ovl_matches_upper(struct dentry *dentry, struct dentry *upper)
+{
+	return d_inode(ovl_dentry_upper(dentry)) == d_inode(upper);
+}
+
 static int ovl_remove_and_whiteout(struct dentry *dentry, bool is_dir)
 {
 	struct dentry *workdir = ovl_workdir(dentry);
@@ -646,7 +660,7 @@ static int ovl_remove_and_whiteout(struct dentry *dentry, bool is_dir)
 	err = -ESTALE;
 	if ((opaquedir && upper != opaquedir) ||
 	    (!opaquedir && ovl_dentry_upper(dentry) &&
-	     upper != ovl_dentry_upper(dentry))) {
+	     !ovl_matches_upper(dentry, upper))) {
 		goto out_dput_upper;
 	}
 
@@ -707,7 +721,7 @@ static int ovl_remove_upper(struct dentry *dentry, bool is_dir)
 
 	err = -ESTALE;
 	if ((opaquedir && upper != opaquedir) ||
-	    (!opaquedir && upper != ovl_dentry_upper(dentry)))
+	    (!opaquedir && !ovl_matches_upper(dentry, upper)))
 		goto out_dput_upper;
 
 	if (is_dir)
@@ -735,8 +749,8 @@ out:
 
 static int ovl_do_remove(struct dentry *dentry, bool is_dir)
 {
-	enum ovl_path_type type;
 	int err;
+	bool locked = false;
 	const struct cred *old_cred;
 
 	err = ovl_want_write(dentry);
@@ -747,7 +761,9 @@ static int ovl_do_remove(struct dentry *dentry, bool is_dir)
 	if (err)
 		goto out_drop_write;
 
-	type = ovl_path_type(dentry);
+	err = ovl_nlink_start(dentry, &locked);
+	if (err)
+		goto out_drop_write;
 
 	old_cred = ovl_override_creds(dentry->d_sb);
 	if (!ovl_lower_positive(dentry))
@@ -761,6 +777,7 @@ static int ovl_do_remove(struct dentry *dentry, bool is_dir)
 		else
 			drop_nlink(dentry->d_inode);
 	}
+	ovl_nlink_end(dentry, locked);
 out_drop_write:
 	ovl_drop_write(dentry);
 out:
@@ -883,6 +900,7 @@ static int ovl_rename(struct inode *olddir, struct dentry *old,
 		      unsigned int flags)
 {
 	int err;
+	bool locked = false;
 	struct dentry *old_upperdir;
 	struct dentry *new_upperdir;
 	struct dentry *olddentry;
@@ -926,6 +944,10 @@ static int ovl_rename(struct inode *olddir, struct dentry *old,
 		err = ovl_copy_up(new);
 		if (err)
 			goto out_drop_write;
+	} else {
+		err = ovl_nlink_start(new, &locked);
+		if (err)
+			goto out_drop_write;
 	}
 
 	old_cred = ovl_override_creds(old->d_sb);
@@ -985,7 +1007,7 @@ static int ovl_rename(struct inode *olddir, struct dentry *old,
 		goto out_unlock;
 
 	err = -ESTALE;
-	if (olddentry != ovl_dentry_upper(old))
+	if (!ovl_matches_upper(old, olddentry))
 		goto out_dput_old;
 
 	newdentry = lookup_one_len(new->d_name.name, new_upperdir,
@@ -998,12 +1020,12 @@ static int ovl_rename(struct inode *olddir, struct dentry *old,
 	new_opaque = ovl_dentry_is_opaque(new);
 
 	err = -ESTALE;
-	if (ovl_dentry_upper(new)) {
+	if (d_inode(new) && ovl_dentry_upper(new)) {
 		if (opaquedir) {
 			if (newdentry != opaquedir)
 				goto out_dput;
 		} else {
-			if (newdentry != ovl_dentry_upper(new))
+			if (!ovl_matches_upper(new, newdentry))
 				goto out_dput;
 		}
 	} else {
@@ -1046,6 +1068,13 @@ static int ovl_rename(struct inode *olddir, struct dentry *old,
 	if (cleanup_whiteout)
 		ovl_cleanup(old_upperdir->d_inode, newdentry);
 
+	if (overwrite && d_inode(new)) {
+		if (new_is_dir)
+			clear_nlink(d_inode(new));
+		else
+			drop_nlink(d_inode(new));
+	}
+
 	ovl_dentry_version_inc(old->d_parent);
 	ovl_dentry_version_inc(new->d_parent);
 
@@ -1057,6 +1086,7 @@ out_unlock:
 	unlock_rename(new_upperdir, old_upperdir);
 out_revert_creds:
 	revert_creds(old_cred);
+	ovl_nlink_end(new, locked);
 out_drop_write:
 	ovl_drop_write(old);
 out:

fs/overlayfs/inode.c

@@ -12,6 +12,7 @@
 #include <linux/cred.h>
 #include <linux/xattr.h>
 #include <linux/posix_acl.h>
+#include <linux/ratelimit.h>
 #include "overlayfs.h"
 
 int ovl_setattr(struct dentry *dentry, struct iattr *attr)
@@ -96,11 +97,15 @@ int ovl_getattr(const struct path *path, struct kstat *stat,
 
 			WARN_ON_ONCE(stat->dev != lowerstat.dev);
 			/*
-			 * Lower hardlinks are broken on copy up to different
+			 * Lower hardlinks may be broken on copy up to different
 			 * upper files, so we cannot use the lower origin st_ino
 			 * for those different files, even for the same fs case.
+			 * With inodes index enabled, it is safe to use st_ino
+			 * of an indexed hardlinked origin. The index validates
+			 * that the upper hardlink is not broken.
 			 */
-			if (is_dir || lowerstat.nlink == 1)
+			if (is_dir || lowerstat.nlink == 1 ||
+			    ovl_test_flag(OVL_INDEX, d_inode(dentry)))
 				stat->ino = lowerstat.ino;
 		}
 		stat->dev = dentry->d_sb->s_dev;
@@ -126,6 +131,15 @@ int ovl_getattr(const struct path *path, struct kstat *stat,
 	if (is_dir && OVL_TYPE_MERGE(type))
 		stat->nlink = 1;
 
+	/*
+	 * Return the overlay inode nlinks for indexed upper inodes.
+	 * Overlay inode nlink counts the union of the upper hardlinks
+	 * and non-covered lower hardlinks. It does not include the upper
+	 * index hardlink.
+	 */
+	if (!is_dir && ovl_test_flag(OVL_INDEX, d_inode(dentry)))
+		stat->nlink = dentry->d_inode->i_nlink;
+
 out:
 	revert_creds(old_cred);
 
@@ -134,8 +148,8 @@ out:
 
 int ovl_permission(struct inode *inode, int mask)
 {
-	bool is_upper;
-	struct inode *realinode = ovl_inode_real(inode, &is_upper);
+	struct inode *upperinode = ovl_inode_upper(inode);
+	struct inode *realinode = upperinode ?: ovl_inode_lower(inode);
 	const struct cred *old_cred;
 	int err;
 
@@ -154,7 +168,8 @@ int ovl_permission(struct inode *inode, int mask)
 		return err;
 
 	old_cred = ovl_override_creds(inode->i_sb);
-	if (!is_upper && !special_file(realinode->i_mode) && mask & MAY_WRITE) {
+	if (!upperinode &&
+	    !special_file(realinode->i_mode) && mask & MAY_WRITE) {
 		mask &= ~(MAY_WRITE | MAY_APPEND);
 		/* Make sure mounter can read file for copy up later */
 		mask |= MAY_READ;
@@ -286,7 +301,7 @@ ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size)
 
 struct posix_acl *ovl_get_acl(struct inode *inode, int type)
 {
-	struct inode *realinode = ovl_inode_real(inode, NULL);
+	struct inode *realinode = ovl_inode_real(inode);
 	const struct cred *old_cred;
 	struct posix_acl *acl;
 
@@ -300,13 +315,13 @@ struct posix_acl *ovl_get_acl(struct inode *inode, int type)
 	return acl;
 }
 
-static bool ovl_open_need_copy_up(int flags, enum ovl_path_type type,
-				  struct dentry *realdentry)
+static bool ovl_open_need_copy_up(struct dentry *dentry, int flags)
 {
-	if (OVL_TYPE_UPPER(type))
+	if (ovl_dentry_upper(dentry) &&
+	    ovl_dentry_has_upper_alias(dentry))
 		return false;
 
-	if (special_file(realdentry->d_inode->i_mode))
+	if (special_file(d_inode(dentry)->i_mode))
 		return false;
 
 	if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC))
@@ -318,11 +333,8 @@ static bool ovl_open_need_copy_up(int flags, enum ovl_path_type type,
 int ovl_open_maybe_copy_up(struct dentry *dentry, unsigned int file_flags)
 {
 	int err = 0;
-	struct path realpath;
-	enum ovl_path_type type;
 
-	type = ovl_path_real(dentry, &realpath);
-	if (ovl_open_need_copy_up(file_flags, type, realpath.dentry)) {
+	if (ovl_open_need_copy_up(dentry, file_flags)) {
 		err = ovl_want_write(dentry);
 		if (!err) {
 			err = ovl_copy_up_flags(dentry, file_flags);
@@ -440,6 +452,103 @@ static void ovl_fill_inode(struct inode *inode, umode_t mode, dev_t rdev)
 	}
 }
 
+/*
+ * With inodes index enabled, an overlay inode nlink counts the union of upper
+ * hardlinks and non-covered lower hardlinks. During the lifetime of a non-pure
+ * upper inode, the following nlink modifying operations can happen:
+ *
+ * 1. Lower hardlink copy up
+ * 2. Upper hardlink created, unlinked or renamed over
+ * 3. Lower hardlink whiteout or renamed over
+ *
+ * For the first, copy up case, the union nlink does not change, whether the
+ * operation succeeds or fails, but the upper inode nlink may change.
+ * Therefore, before copy up, we store the union nlink value relative to the
+ * lower inode nlink in the index inode xattr trusted.overlay.nlink.
+ *
+ * For the second, upper hardlink case, the union nlink should be incremented
+ * or decremented IFF the operation succeeds, aligned with nlink change of the
+ * upper inode. Therefore, before link/unlink/rename, we store the union nlink
+ * value relative to the upper inode nlink in the index inode.
+ *
+ * For the last, lower cover up case, we simplify things by preceding the
+ * whiteout or cover up with copy up. This makes sure that there is an index
+ * upper inode where the nlink xattr can be stored before the copied up upper
+ * entry is unlink.
+ */
+#define OVL_NLINK_ADD_UPPER	(1 << 0)
+
+/*
+ * On-disk format for indexed nlink:
+ *
+ * nlink relative to the upper inode - "U[+-]NUM"
+ * nlink relative to the lower inode - "L[+-]NUM"
+ */
+
+static int ovl_set_nlink_common(struct dentry *dentry,
+				struct dentry *realdentry, const char *format)
+{
+	struct inode *inode = d_inode(dentry);
+	struct inode *realinode = d_inode(realdentry);
+	char buf[13];
+	int len;
+
+	len = snprintf(buf, sizeof(buf), format,
+		       (int) (inode->i_nlink - realinode->i_nlink));
+
+	return ovl_do_setxattr(ovl_dentry_upper(dentry),
+			       OVL_XATTR_NLINK, buf, len, 0);
+}
+
+int ovl_set_nlink_upper(struct dentry *dentry)
+{
+	return ovl_set_nlink_common(dentry, ovl_dentry_upper(dentry), "U%+i");
+}
+
+int ovl_set_nlink_lower(struct dentry *dentry)
+{
+	return ovl_set_nlink_common(dentry, ovl_dentry_lower(dentry), "L%+i");
+}
+
+unsigned int ovl_get_nlink(struct dentry *lowerdentry,
+			   struct dentry *upperdentry,
+			   unsigned int fallback)
+{
+	int nlink_diff;
+	int nlink;
+	char buf[13];
+	int err;
+
+	if (!lowerdentry || !upperdentry || d_inode(lowerdentry)->i_nlink == 1)
+		return fallback;
+
+	err = vfs_getxattr(upperdentry, OVL_XATTR_NLINK, &buf, sizeof(buf) - 1);
+	if (err < 0)
+		goto fail;
+
+	buf[err] = '\0';
+	if ((buf[0] != 'L' && buf[0] != 'U') ||
+	    (buf[1] != '+' && buf[1] != '-'))
+		goto fail;
+
+	err = kstrtoint(buf + 1, 10, &nlink_diff);
+	if (err < 0)
+		goto fail;
+
+	nlink = d_inode(buf[0] == 'L' ? lowerdentry : upperdentry)->i_nlink;
+	nlink += nlink_diff;
+
+	if (nlink <= 0)
+		goto fail;
+
+	return nlink;
+
+fail:
+	pr_warn_ratelimited("overlayfs: failed to get index nlink (%pd2, err=%i)\n",
+			    upperdentry, err);
+	return fallback;
+}
+
 struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, dev_t rdev)
 {
 	struct inode *inode;
@@ -453,27 +562,87 @@ struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, dev_t rdev)
 
 static int ovl_inode_test(struct inode *inode, void *data)
 {
-	return ovl_inode_real(inode, NULL) == data;
+	return inode->i_private == data;
 }
 
 static int ovl_inode_set(struct inode *inode, void *data)
 {
-	inode->i_private = (void *) (((unsigned long) data) | OVL_ISUPPER_MASK);
+	inode->i_private = data;
 	return 0;
 }
 
-struct inode *ovl_get_inode(struct super_block *sb, struct inode *realinode)
+static bool ovl_verify_inode(struct inode *inode, struct dentry *lowerdentry,
+			     struct dentry *upperdentry)
+{
+	struct inode *lowerinode = lowerdentry ? d_inode(lowerdentry) : NULL;
+
+	/* Lower (origin) inode must match, even if NULL */
+	if (ovl_inode_lower(inode) != lowerinode)
+		return false;
+
+	/*
+	 * Allow non-NULL __upperdentry in inode even if upperdentry is NULL.
+	 * This happens when finding a lower alias for a copied up hard link.
+	 */
+	if (upperdentry && ovl_inode_upper(inode) != d_inode(upperdentry))
+		return false;
 
+	return true;
+}
+
+struct inode *ovl_get_inode(struct dentry *dentry, struct dentry *upperdentry)
 {
+	struct dentry *lowerdentry = ovl_dentry_lower(dentry);
+	struct inode *realinode = upperdentry ? d_inode(upperdentry) : NULL;
 	struct inode *inode;
 
-	inode = iget5_locked(sb, (unsigned long) realinode,
-			     ovl_inode_test, ovl_inode_set, realinode);
-	if (inode && inode->i_state & I_NEW) {
-		ovl_fill_inode(inode, realinode->i_mode, realinode->i_rdev);
-		set_nlink(inode, realinode->i_nlink);
-		unlock_new_inode(inode);
+	if (!realinode)
+		realinode = d_inode(lowerdentry);
+
+	if (!S_ISDIR(realinode->i_mode) &&
+	    (upperdentry || (lowerdentry && ovl_indexdir(dentry->d_sb)))) {
+		struct inode *key = d_inode(lowerdentry ?: upperdentry);
+		unsigned int nlink;
+
+		inode = iget5_locked(dentry->d_sb, (unsigned long) key,
+				     ovl_inode_test, ovl_inode_set, key);
+		if (!inode)
+			goto out_nomem;
+		if (!(inode->i_state & I_NEW)) {
+			/*
+			 * Verify that the underlying files stored in the inode
+			 * match those in the dentry.
+			 */
+			if (!ovl_verify_inode(inode, lowerdentry, upperdentry)) {
+				iput(inode);
+				inode = ERR_PTR(-ESTALE);
+				goto out;
+			}
+
+			dput(upperdentry);
+			goto out;
+		}
+
+		nlink = ovl_get_nlink(lowerdentry, upperdentry,
+				      realinode->i_nlink);
+		set_nlink(inode, nlink);
+	} else {
+		inode = new_inode(dentry->d_sb);
+		if (!inode)
+			goto out_nomem;
 	}
+	ovl_fill_inode(inode, realinode->i_mode, realinode->i_rdev);
+	ovl_inode_init(inode, upperdentry, lowerdentry);
+
+	if (upperdentry && ovl_is_impuredir(upperdentry))
+		ovl_set_flag(OVL_IMPURE, inode);
 
+	if (inode->i_state & I_NEW)
+		unlock_new_inode(inode);
+out:
 	return inode;
+
+out_nomem:
+	inode = ERR_PTR(-ENOMEM);
+	goto out;
 }

fs/overlayfs/namei.c

@@ -88,13 +88,10 @@ static int ovl_acceptable(void *ctx, struct dentry *dentry)
 	return 1;
 }
 
-static struct dentry *ovl_get_origin(struct dentry *dentry,
-				     struct vfsmount *mnt)
+static struct ovl_fh *ovl_get_origin_fh(struct dentry *dentry)
 {
 	int res;
 	struct ovl_fh *fh = NULL;
-	struct dentry *origin = NULL;
-	int bytes;
 
 	res = vfs_getxattr(dentry, OVL_XATTR_ORIGIN, NULL, 0);
 	if (res < 0) {
@@ -106,7 +103,7 @@ static struct dentry *ovl_get_origin(struct dentry *dentry,
 	if (res == 0)
 		return NULL;
 
-	fh  = kzalloc(res, GFP_TEMPORARY);
+	fh = kzalloc(res, GFP_TEMPORARY);
 	if (!fh)
 		return ERR_PTR(-ENOMEM);
 
@@ -129,7 +126,29 @@ static struct dentry *ovl_get_origin(struct dentry *dentry,
 	    (fh->flags & OVL_FH_FLAG_BIG_ENDIAN) != OVL_FH_FLAG_CPU_ENDIAN)
 		goto out;
 
-	bytes = (fh->len - offsetof(struct ovl_fh, fid));
+	return fh;
+
+out:
+	kfree(fh);
+	return NULL;
+
+fail:
+	pr_warn_ratelimited("overlayfs: failed to get origin (%i)\n", res);
+	goto out;
+invalid:
+	pr_warn_ratelimited("overlayfs: invalid origin (%*phN)\n", res, fh);
+	goto out;
+}
+
+static struct dentry *ovl_get_origin(struct dentry *dentry,
+				     struct vfsmount *mnt)
+{
+	struct dentry *origin = NULL;
+	struct ovl_fh *fh = ovl_get_origin_fh(dentry);
+	int bytes;
+
+	if (IS_ERR_OR_NULL(fh))
+		return (struct dentry *)fh;
 
 	/*
 	 * Make sure that the stored uuid matches the uuid of the lower
@@ -138,6 +157,7 @@ static struct dentry *ovl_get_origin(struct dentry *dentry,
 	if (!uuid_equal(&fh->uuid, &mnt->mnt_sb->s_uuid))
 		goto out;
 
+	bytes = (fh->len - offsetof(struct ovl_fh, fid));
 	origin = exportfs_decode_fh(mnt, (struct fid *)fh->fid,
 				    bytes >> 2, (int)fh->type,
 				    ovl_acceptable, NULL);
@@ -149,21 +169,17 @@ static struct dentry *ovl_get_origin(struct dentry *dentry,
 	}
 
 	if (ovl_dentry_weird(origin) ||
-	    ((d_inode(origin)->i_mode ^ d_inode(dentry)->i_mode) & S_IFMT)) {
-		dput(origin);
-		origin = NULL;
+	    ((d_inode(origin)->i_mode ^ d_inode(dentry)->i_mode) & S_IFMT))
 		goto invalid;
-	}
 
 out:
 	kfree(fh);
 	return origin;
 
-fail:
-	pr_warn_ratelimited("overlayfs: failed to get origin (%i)\n", res);
-	goto out;
 invalid:
-	pr_warn_ratelimited("overlayfs: invalid origin (%*phN)\n", res, fh);
+	pr_warn_ratelimited("overlayfs: invalid origin (%pd2)\n", origin);
+	dput(origin);
+	origin = NULL;
 	goto out;
 }
 
@@ -269,34 +285,31 @@ static int ovl_lookup_layer(struct dentry *base, struct ovl_lookup_data *d,
 }
 
 
-static int ovl_check_origin(struct dentry *dentry, struct dentry *upperdentry,
+static int ovl_check_origin(struct dentry *upperdentry,
+			    struct path *lowerstack, unsigned int numlower,
 			    struct path **stackp, unsigned int *ctrp)
 {
-	struct super_block *same_sb = ovl_same_sb(dentry->d_sb);
-	struct ovl_entry *roe = dentry->d_sb->s_root->d_fsdata;
 	struct vfsmount *mnt;
-	struct dentry *origin;
+	struct dentry *origin = NULL;
+	int i;
+
+
+	for (i = 0; i < numlower; i++) {
+		mnt = lowerstack[i].mnt;
+		origin = ovl_get_origin(upperdentry, mnt);
+		if (IS_ERR(origin))
+			return PTR_ERR(origin);
 
-	if (!same_sb || !roe->numlower)
+		if (origin)
+			break;
+	}
+
+	if (!origin)
 		return 0;
 
-       /*
-	* Since all layers are on the same fs, we use the first layer for
-	* decoding the file handle.  We may get a disconnected dentry,
-	* which is fine, because we only need to hold the origin inode in
-	* cache and use its inode number.  We may even get a connected dentry,
-	* that is not under the first layer's root.  That is also fine for
-	* using it's inode number - it's the same as if we held a reference
-	* to a dentry in first layer that was moved under us.
-	*/
-	mnt = roe->lowerstack[0].mnt;
-
-	origin = ovl_get_origin(upperdentry, mnt);
-	if (IS_ERR_OR_NULL(origin))
-		return PTR_ERR(origin);
-
-	BUG_ON(*stackp || *ctrp);
-	*stackp = kmalloc(sizeof(struct path), GFP_TEMPORARY);
+	BUG_ON(*ctrp);
+	if (!*stackp)
+		*stackp = kmalloc(sizeof(struct path), GFP_TEMPORARY);
 	if (!*stackp) {
 		dput(origin);
 		return -ENOMEM;
@@ -307,6 +320,215 @@ static int ovl_check_origin(struct dentry *dentry, struct dentry *upperdentry,
 	return 0;
 }
 
+/*
+ * Verify that @fh matches the origin file handle stored in OVL_XATTR_ORIGIN.
+ * Return 0 on match, -ESTALE on mismatch, < 0 on error.
+ */
+static int ovl_verify_origin_fh(struct dentry *dentry, const struct ovl_fh *fh)
+{
+	struct ovl_fh *ofh = ovl_get_origin_fh(dentry);
+	int err = 0;
+
+	if (!ofh)
+		return -ENODATA;
+
+	if (IS_ERR(ofh))
+		return PTR_ERR(ofh);
+
+	if (fh->len != ofh->len || memcmp(fh, ofh, fh->len))
+		err = -ESTALE;
+
+	kfree(ofh);
+	return err;
+}
+
+/*
+ * Verify that an inode matches the origin file handle stored in upper inode.
+ *
+ * If @set is true and there is no stored file handle, encode and store origin
+ * file handle in OVL_XATTR_ORIGIN.
+ *
+ * Return 0 on match, -ESTALE on mismatch, < 0 on error.
+ */
+int ovl_verify_origin(struct dentry *dentry, struct vfsmount *mnt,
+		      struct dentry *origin, bool is_upper, bool set)
+{
+	struct inode *inode;
+	struct ovl_fh *fh;
+	int err;
+
+	fh = ovl_encode_fh(origin, is_upper);
+	err = PTR_ERR(fh);
+	if (IS_ERR(fh))
+		goto fail;
+
+	err = ovl_verify_origin_fh(dentry, fh);
+	if (set && err == -ENODATA)
+		err = ovl_do_setxattr(dentry, OVL_XATTR_ORIGIN, fh, fh->len, 0);
+	if (err)
+		goto fail;
+
+out:
+	kfree(fh);
+	return err;
+
+fail:
+	inode = d_inode(origin);
+	pr_warn_ratelimited("overlayfs: failed to verify origin (%pd2, ino=%lu, err=%i)\n",
+			    origin, inode ? inode->i_ino : 0, err);
+	goto out;
+}
+
+/*
+ * Verify that an index entry name matches the origin file handle stored in
+ * OVL_XATTR_ORIGIN and that origin file handle can be decoded to lower path.
+ * Return 0 on match, -ESTALE on mismatch or stale origin, < 0 on error.
+ */
+int ovl_verify_index(struct dentry *index, struct path *lowerstack,
+		     unsigned int numlower)
+{
+	struct ovl_fh *fh = NULL;
+	size_t len;
+	struct path origin = { };
+	struct path *stack = &origin;
+	unsigned int ctr = 0;
+	int err;
+
+	if (!d_inode(index))
+		return 0;
+
+	err = -EISDIR;
+	if (d_is_dir(index))
+		goto fail;
+
+	err = -EINVAL;
+	if (index->d_name.len < sizeof(struct ovl_fh)*2)
+		goto fail;
+
+	err = -ENOMEM;
+	len = index->d_name.len / 2;
+	fh = kzalloc(len, GFP_TEMPORARY);
+	if (!fh)
+		goto fail;
+
+	err = -EINVAL;
+	if (hex2bin((u8 *)fh, index->d_name.name, len) || len != fh->len)
+		goto fail;
+
+	err = ovl_verify_origin_fh(index, fh);
+	if (err)
+		goto fail;
+
+	err = ovl_check_origin(index, lowerstack, numlower, &stack, &ctr);
+	if (!err && !ctr)
+		err = -ESTALE;
+	if (err)
+		goto fail;
+
+	/* Check if index is orphan and don't warn before cleaning it */
+	if (d_inode(index)->i_nlink == 1 &&
+	    ovl_get_nlink(index, origin.dentry, 0) == 0)
+		err = -ENOENT;
+
+	dput(origin.dentry);
+out:
+	kfree(fh);
+	return err;
+
+fail:
+	pr_warn_ratelimited("overlayfs: failed to verify index (%pd2, err=%i)\n",
+			    index, err);
+	goto out;
+}
+
+/*
+ * Lookup in indexdir for the index entry of a lower real inode or a copy up
+ * origin inode. The index entry name is the hex representation of the lower
+ * inode file handle.
+ *
+ * If the index dentry in negative, then either no lower aliases have been
+ * copied up yet, or aliases have been copied up in older kernels and are
+ * not indexed.
+ *
+ * If the index dentry for a copy up origin inode is positive, but points
+ * to an inode different than the upper inode, then either the upper inode
+ * has been copied up and not indexed or it was indexed, but since then
+ * index dir was cleared. Either way, that index cannot be used to indentify
+ * the overlay inode.
+ */
+int ovl_get_index_name(struct dentry *origin, struct qstr *name)
+{
+	int err;
+	struct ovl_fh *fh;
+	char *n, *s;
+
+	fh = ovl_encode_fh(origin, false);
+	if (IS_ERR(fh))
+		return PTR_ERR(fh);
+
+	err = -ENOMEM;
+	n = kzalloc(fh->len * 2, GFP_TEMPORARY);
+	if (n) {
+		s  = bin2hex(n, fh, fh->len);
+		*name = (struct qstr) QSTR_INIT(n, s - n);
+		err = 0;
+	}
+	kfree(fh);
+
+	return err;
+
+}
+
+static struct dentry *ovl_lookup_index(struct dentry *dentry,
+				       struct dentry *upper,
+				       struct dentry *origin)
+{
+	struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
+	struct dentry *index;
+	struct inode *inode;
+	struct qstr name;
+	int err;
+
+	err = ovl_get_index_name(origin, &name);
+	if (err)
+		return ERR_PTR(err);
+
+	index = lookup_one_len_unlocked(name.name, ofs->indexdir, name.len);
+	if (IS_ERR(index)) {
+		pr_warn_ratelimited("overlayfs: failed inode index lookup (ino=%lu, key=%*s, err=%i);\n"
+				    "overlayfs: mount with '-o index=off' to disable inodes index.\n",
+				    d_inode(origin)->i_ino, name.len, name.name,
+				    err);
+		goto out;
+	}
+
+	if (d_is_negative(index)) {
+		if (upper && d_inode(origin)->i_nlink > 1) {
+			pr_warn_ratelimited("overlayfs: hard link with origin but no index (ino=%lu).\n",
+					    d_inode(origin)->i_ino);
+			goto fail;
+		}
+
+		dput(index);
+		index = NULL;
+	} else if (upper && d_inode(index) != d_inode(upper)) {
+		inode = d_inode(index);
+		pr_warn_ratelimited("overlayfs: wrong index found (index ino: %lu, upper ino: %lu).\n",
+				    d_inode(index)->i_ino,
+				    d_inode(upper)->i_ino);
+		goto fail;
+	}
+
+out:
+	kfree(name.name);
+	return index;
+
+fail:
+	dput(index);
+	index = ERR_PTR(-EIO);
+	goto out;
+}
+
 /*
  * Returns next layer in stack starting from top.
  * Returns -1 if this is the last layer.
@@ -338,10 +560,10 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 	struct ovl_entry *roe = dentry->d_sb->s_root->d_fsdata;
 	struct path *stack = NULL;
 	struct dentry *upperdir, *upperdentry = NULL;
+	struct dentry *index = NULL;
 	unsigned int ctr = 0;
 	struct inode *inode = NULL;
 	bool upperopaque = false;
-	bool upperimpure = false;
 	char *upperredirect = NULL;
 	struct dentry *this;
 	unsigned int i;
@@ -359,7 +581,7 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 		return ERR_PTR(-ENAMETOOLONG);
 
 	old_cred = ovl_override_creds(dentry->d_sb);
-	upperdir = ovl_upperdentry_dereference(poe);
+	upperdir = ovl_dentry_upper(dentry->d_parent);
 	if (upperdir) {
 		err = ovl_lookup_layer(upperdir, &d, &upperdentry);
 		if (err)
@@ -372,8 +594,18 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 		}
 		if (upperdentry && !d.is_dir) {
 			BUG_ON(!d.stop || d.redirect);
-			err = ovl_check_origin(dentry, upperdentry,
-					       &stack, &ctr);
+			/*
+			 * Lookup copy up origin by decoding origin file handle.
+			 * We may get a disconnected dentry, which is fine,
+			 * because we only need to hold the origin inode in
+			 * cache and use its inode number.  We may even get a
+			 * connected dentry, that is not under any of the lower
+			 * layers root.  That is also fine for using it's inode
+			 * number - it's the same as if we held a reference
+			 * to a dentry in lower layer that was moved under us.
+			 */
+			err = ovl_check_origin(upperdentry, roe->lowerstack,
+					       roe->numlower, &stack, &ctr);
 			if (err)
 				goto out;
 		}
@@ -386,8 +618,6 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 				poe = roe;
 		}
 		upperopaque = d.opaque;
-		if (upperdentry && d.is_dir)
-			upperimpure = ovl_is_impuredir(upperdentry);
 	}
 
 	if (!d.stop && poe->numlower) {
@@ -428,48 +658,56 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 		}
 	}
 
+	/* Lookup index by lower inode and verify it matches upper inode */
+	if (ctr && !d.is_dir && ovl_indexdir(dentry->d_sb)) {
+		struct dentry *origin = stack[0].dentry;
+
+		index = ovl_lookup_index(dentry, upperdentry, origin);
+		if (IS_ERR(index)) {
+			err = PTR_ERR(index);
+			index = NULL;
+			goto out_put;
+		}
+	}
+
 	oe = ovl_alloc_entry(ctr);
 	err = -ENOMEM;
 	if (!oe)
 		goto out_put;
 
-	if (upperdentry || ctr) {
-		struct dentry *realdentry;
-		struct inode *realinode;
+	oe->opaque = upperopaque;
+	memcpy(oe->lowerstack, stack, sizeof(struct path) * ctr);
+	dentry->d_fsdata = oe;
 
-		realdentry = upperdentry ? upperdentry : stack[0].dentry;
-		realinode = d_inode(realdentry);
+	if (upperdentry)
+		ovl_dentry_set_upper_alias(dentry);
+	else if (index)
+		upperdentry = dget(index);
 
-		err = -ENOMEM;
-		if (upperdentry && !d_is_dir(upperdentry)) {
-			inode = ovl_get_inode(dentry->d_sb, realinode);
-		} else {
-			inode = ovl_new_inode(dentry->d_sb, realinode->i_mode,
-					      realinode->i_rdev);
-			if (inode)
-				ovl_inode_init(inode, realinode, !!upperdentry);
-		}
-		if (!inode)
+	if (upperdentry || ctr) {
+		inode = ovl_get_inode(dentry, upperdentry);
+		err = PTR_ERR(inode);
+		if (IS_ERR(inode))
 			goto out_free_oe;
-		ovl_copyattr(realdentry->d_inode, inode);
+
+		OVL_I(inode)->redirect = upperredirect;
+		if (index)
+			ovl_set_flag(OVL_INDEX, inode);
 	}
 
 	revert_creds(old_cred);
-	oe->opaque = upperopaque;
-	oe->impure = upperimpure;
-	oe->redirect = upperredirect;
-	oe->__upperdentry = upperdentry;
-	memcpy(oe->lowerstack, stack, sizeof(struct path) * ctr);
+	dput(index);
 	kfree(stack);
 	kfree(d.redirect);
-	dentry->d_fsdata = oe;
 	d_add(dentry, inode);
 
 	return NULL;
 
 out_free_oe:
+	dentry->d_fsdata = NULL;
 	kfree(oe);
 out_put:
+	dput(index);
 	for (i = 0; i < ctr; i++)
 		dput(stack[i].dentry);
 	kfree(stack);
@@ -499,7 +737,7 @@ bool ovl_lower_positive(struct dentry *dentry)
 		return oe->opaque;
 
 	/* Negative upper -> positive lower */
-	if (!oe->__upperdentry)
+	if (!ovl_dentry_upper(dentry))
 		return true;
 
 	/* Positive upper -> have to look up lower to see whether it exists */

fs/overlayfs/overlayfs.h

@@ -25,6 +25,12 @@ enum ovl_path_type {
 #define OVL_XATTR_REDIRECT OVL_XATTR_PREFIX "redirect"
 #define OVL_XATTR_ORIGIN OVL_XATTR_PREFIX "origin"
 #define OVL_XATTR_IMPURE OVL_XATTR_PREFIX "impure"
+#define OVL_XATTR_NLINK OVL_XATTR_PREFIX "nlink"
+
+enum ovl_flag {
+	OVL_IMPURE,
+	OVL_INDEX,
+};
 
 /*
  * The tuple (fh,uuid) is a universal unique identifier for a copy up origin,
@@ -38,6 +44,8 @@ enum ovl_path_type {
 /* CPU byte order required for fid decoding:  */
 #define OVL_FH_FLAG_BIG_ENDIAN	(1 << 0)
 #define OVL_FH_FLAG_ANY_ENDIAN	(1 << 1)
+/* Is the real inode encoded in fid an upper inode? */
+#define OVL_FH_FLAG_PATH_UPPER	(1 << 2)
 
 #define OVL_FH_FLAG_ALL (OVL_FH_FLAG_BIG_ENDIAN | OVL_FH_FLAG_ANY_ENDIAN)
 
@@ -60,8 +68,6 @@ struct ovl_fh {
 	u8 fid[0];	/* file identifier */
 } __packed;
 
-#define OVL_ISUPPER_MASK 1UL
-
 static inline int ovl_do_rmdir(struct inode *dir, struct dentry *dentry)
 {
 	int err = vfs_rmdir(dir, dentry);
@@ -175,22 +181,14 @@ static inline struct dentry *ovl_do_tmpfile(struct dentry *dentry, umode_t mode)
 	return ret;
 }
 
-static inline struct inode *ovl_inode_real(struct inode *inode, bool *is_upper)
-{
-	unsigned long x = (unsigned long) READ_ONCE(inode->i_private);
-
-	if (is_upper)
-		*is_upper = x & OVL_ISUPPER_MASK;
-
-	return (struct inode *) (x & ~OVL_ISUPPER_MASK);
-}
-
 /* util.c */
 int ovl_want_write(struct dentry *dentry);
 void ovl_drop_write(struct dentry *dentry);
 struct dentry *ovl_workdir(struct dentry *dentry);
 const struct cred *ovl_override_creds(struct super_block *sb);
 struct super_block *ovl_same_sb(struct super_block *sb);
+bool ovl_can_decode_fh(struct super_block *sb);
+struct dentry *ovl_indexdir(struct super_block *sb);
 struct ovl_entry *ovl_alloc_entry(unsigned int numlower);
 bool ovl_dentry_remote(struct dentry *dentry);
 bool ovl_dentry_weird(struct dentry *dentry);
@@ -201,19 +199,22 @@ enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path);
 struct dentry *ovl_dentry_upper(struct dentry *dentry);
 struct dentry *ovl_dentry_lower(struct dentry *dentry);
 struct dentry *ovl_dentry_real(struct dentry *dentry);
+struct inode *ovl_inode_upper(struct inode *inode);
+struct inode *ovl_inode_lower(struct inode *inode);
+struct inode *ovl_inode_real(struct inode *inode);
 struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry);
 void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache);
 bool ovl_dentry_is_opaque(struct dentry *dentry);
-bool ovl_dentry_is_impure(struct dentry *dentry);
 bool ovl_dentry_is_whiteout(struct dentry *dentry);
 void ovl_dentry_set_opaque(struct dentry *dentry);
+bool ovl_dentry_has_upper_alias(struct dentry *dentry);
+void ovl_dentry_set_upper_alias(struct dentry *dentry);
 bool ovl_redirect_dir(struct super_block *sb);
 const char *ovl_dentry_get_redirect(struct dentry *dentry);
 void ovl_dentry_set_redirect(struct dentry *dentry, const char *redirect);
-void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry);
-void ovl_inode_init(struct inode *inode, struct inode *realinode,
-		    bool is_upper);
-void ovl_inode_update(struct inode *inode, struct inode *upperinode);
+void ovl_inode_init(struct inode *inode, struct dentry *upperdentry,
+		    struct dentry *lowerdentry);
+void ovl_inode_update(struct inode *inode, struct dentry *upperdentry);
 void ovl_dentry_version_inc(struct dentry *dentry);
 u64 ovl_dentry_version_get(struct dentry *dentry);
 bool ovl_is_whiteout(struct dentry *dentry);
@@ -225,6 +226,12 @@ int ovl_check_setxattr(struct dentry *dentry, struct dentry *upperdentry,
 		       const char *name, const void *value, size_t size,
 		       int xerr);
 int ovl_set_impure(struct dentry *dentry, struct dentry *upperdentry);
+void ovl_set_flag(unsigned long flag, struct inode *inode);
+bool ovl_test_flag(unsigned long flag, struct inode *inode);
+bool ovl_inuse_trylock(struct dentry *dentry);
+void ovl_inuse_unlock(struct dentry *dentry);
+int ovl_nlink_start(struct dentry *dentry, bool *locked);
+void ovl_nlink_end(struct dentry *dentry, bool locked);
 
 static inline bool ovl_is_impuredir(struct dentry *dentry)
 {
@@ -233,6 +240,11 @@ static inline bool ovl_is_impuredir(struct dentry *dentry)
 
 
 /* namei.c */
+int ovl_verify_origin(struct dentry *dentry, struct vfsmount *mnt,
+		      struct dentry *origin, bool is_upper, bool set);
+int ovl_verify_index(struct dentry *index, struct path *lowerstack,
+		     unsigned int numlower);
+int ovl_get_index_name(struct dentry *origin, struct qstr *name);
 int ovl_path_next(int idx, struct dentry *dentry, struct path *path);
 struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags);
 bool ovl_lower_positive(struct dentry *dentry);
@@ -245,8 +257,15 @@ void ovl_cache_free(struct list_head *list);
 int ovl_check_d_type_supported(struct path *realpath);
 void ovl_workdir_cleanup(struct inode *dir, struct vfsmount *mnt,
 			 struct dentry *dentry, int level);
+int ovl_indexdir_cleanup(struct dentry *dentry, struct vfsmount *mnt,
+			 struct path *lowerstack, unsigned int numlower);
 
 /* inode.c */
+int ovl_set_nlink_upper(struct dentry *dentry);
+int ovl_set_nlink_lower(struct dentry *dentry);
+unsigned int ovl_get_nlink(struct dentry *lowerdentry,
+			   struct dentry *upperdentry,
+			   unsigned int fallback);
 int ovl_setattr(struct dentry *dentry, struct iattr *attr);
 int ovl_getattr(const struct path *path, struct kstat *stat,
 		u32 request_mask, unsigned int flags);
@@ -262,7 +281,7 @@ int ovl_update_time(struct inode *inode, struct timespec *ts, int flags);
 bool ovl_is_private_xattr(const char *name);
 
 struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, dev_t rdev);
-struct inode *ovl_get_inode(struct super_block *sb, struct inode *realinode);
+struct inode *ovl_get_inode(struct dentry *dentry, struct dentry *upperdentry);
 static inline void ovl_copyattr(struct inode *from, struct inode *to)
 {
 	to->i_uid = from->i_uid;
@@ -284,10 +303,11 @@ struct cattr {
 int ovl_create_real(struct inode *dir, struct dentry *newdentry,
 		    struct cattr *attr,
 		    struct dentry *hardlink, bool debug);
-void ovl_cleanup(struct inode *dir, struct dentry *dentry);
+int ovl_cleanup(struct inode *dir, struct dentry *dentry);
 
 /* copy_up.c */
 int ovl_copy_up(struct dentry *dentry);
 int ovl_copy_up_flags(struct dentry *dentry, int flags);
 int ovl_copy_xattr(struct dentry *old, struct dentry *new);
 int ovl_set_attr(struct dentry *upper, struct kstat *stat);
+struct ovl_fh *ovl_encode_fh(struct dentry *lower, bool is_upper);

fs/overlayfs/ovl_entry.h

@@ -14,6 +14,7 @@ struct ovl_config {
 	char *workdir;
 	bool default_permissions;
 	bool redirect_dir;
+	bool index;
 };
 
 /* private information held for overlayfs's superblock */
@@ -21,7 +22,12 @@ struct ovl_fs {
 	struct vfsmount *upper_mnt;
 	unsigned numlower;
 	struct vfsmount **lower_mnt;
+	/* workbasedir is the path at workdir= mount option */
+	struct dentry *workbasedir;
+	/* workdir is the 'work' directory under workbasedir */
 	struct dentry *workdir;
+	/* index directory listing overlay inodes by origin file handle */
+	struct dentry *indexdir;
 	long namelen;
 	/* pathnames of lower and upper dirs, for show_options */
 	struct ovl_config config;
@@ -29,22 +35,16 @@ struct ovl_fs {
 	const struct cred *creator_cred;
 	bool tmpfile;
 	bool noxattr;
-	wait_queue_head_t copyup_wq;
 	/* sb common to all layers */
 	struct super_block *same_sb;
 };
 
 /* private information held for every overlayfs dentry */
 struct ovl_entry {
-	struct dentry *__upperdentry;
-	struct ovl_dir_cache *cache;
 	union {
 		struct {
-			u64 version;
-			const char *redirect;
+			unsigned long has_upper;
 			bool opaque;
-			bool impure;
-			bool copying;
 		};
 		struct rcu_head rcu;
 	};
@@ -54,7 +54,25 @@ struct ovl_entry {
 
 struct ovl_entry *ovl_alloc_entry(unsigned int numlower);
 
-static inline struct dentry *ovl_upperdentry_dereference(struct ovl_entry *oe)
+struct ovl_inode {
+	struct ovl_dir_cache *cache;
+	const char *redirect;
+	u64 version;
+	unsigned long flags;
+	struct inode vfs_inode;
+	struct dentry *__upperdentry;
+	struct inode *lower;
+
+	/* synchronize copy up and more */
+	struct mutex lock;
+};
+
+static inline struct ovl_inode *OVL_I(struct inode *inode)
+{
+	return container_of(inode, struct ovl_inode, vfs_inode);
+}
+
+static inline struct dentry *ovl_upperdentry_dereference(struct ovl_inode *oi)
 {
-	return lockless_dereference(oe->__upperdentry);
+	return lockless_dereference(oi->__upperdentry);
 }

fs/overlayfs/readdir.c

@@ -667,3 +667,53 @@ void ovl_workdir_cleanup(struct inode *dir, struct vfsmount *mnt,
 		ovl_cleanup(dir, dentry);
 	}
 }
+
+int ovl_indexdir_cleanup(struct dentry *dentry, struct vfsmount *mnt,
+			 struct path *lowerstack, unsigned int numlower)
+{
+	int err;
+	struct inode *dir = dentry->d_inode;
+	struct path path = { .mnt = mnt, .dentry = dentry };
+	LIST_HEAD(list);
+	struct ovl_cache_entry *p;
+	struct ovl_readdir_data rdd = {
+		.ctx.actor = ovl_fill_merge,
+		.dentry = NULL,
+		.list = &list,
+		.root = RB_ROOT,
+		.is_lowest = false,
+	};
+
+	err = ovl_dir_read(&path, &rdd);
+	if (err)
+		goto out;
+
+	inode_lock_nested(dir, I_MUTEX_PARENT);
+	list_for_each_entry(p, &list, l_node) {
+		struct dentry *index;
+
+		if (p->name[0] == '.') {
+			if (p->len == 1)
+				continue;
+			if (p->len == 2 && p->name[1] == '.')
+				continue;
+		}
+		index = lookup_one_len(p->name, dentry, p->len);
+		if (IS_ERR(index)) {
+			err = PTR_ERR(index);
+			break;
+		}
+		if (ovl_verify_index(index, lowerstack, numlower)) {
+			err = ovl_cleanup(dir, index);
+			if (err)
+				break;
+		}
+		dput(index);
+	}
+	inode_unlock(dir);
+out:
+	ovl_cache_free(&list);
+	if (err)
+		pr_err("overlayfs: failed index dir cleanup (%i)\n", err);
+	return err;
+}

fs/overlayfs/super.c

@@ -34,6 +34,11 @@ module_param_named(redirect_dir, ovl_redirect_dir_def, bool, 0644);
 MODULE_PARM_DESC(ovl_redirect_dir_def,
 		 "Default to on or off for the redirect_dir feature");
 
+static bool ovl_index_def = IS_ENABLED(CONFIG_OVERLAY_FS_INDEX);
+module_param_named(index, ovl_index_def, bool, 0644);
+MODULE_PARM_DESC(ovl_index_def,
+		 "Default to on or off for the inodes index feature");
+
 static void ovl_dentry_release(struct dentry *dentry)
 {
 	struct ovl_entry *oe = dentry->d_fsdata;
@@ -41,8 +46,6 @@ static void ovl_dentry_release(struct dentry *dentry)
 	if (oe) {
 		unsigned int i;
 
-		dput(oe->__upperdentry);
-		kfree(oe->redirect);
 		for (i = 0; i < oe->numlower; i++)
 			dput(oe->lowerstack[i].dentry);
 		kfree_rcu(oe, rcu);
@@ -165,12 +168,52 @@ static const struct dentry_operations ovl_reval_dentry_operations = {
 	.d_weak_revalidate = ovl_dentry_weak_revalidate,
 };
 
+static struct kmem_cache *ovl_inode_cachep;
+
+static struct inode *ovl_alloc_inode(struct super_block *sb)
+{
+	struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL);
+
+	oi->cache = NULL;
+	oi->redirect = NULL;
+	oi->version = 0;
+	oi->flags = 0;
+	oi->__upperdentry = NULL;
+	oi->lower = NULL;
+	mutex_init(&oi->lock);
+
+	return &oi->vfs_inode;
+}
+
+static void ovl_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+
+	kmem_cache_free(ovl_inode_cachep, OVL_I(inode));
+}
+
+static void ovl_destroy_inode(struct inode *inode)
+{
+	struct ovl_inode *oi = OVL_I(inode);
+
+	dput(oi->__upperdentry);
+	kfree(oi->redirect);
+	mutex_destroy(&oi->lock);
+
+	call_rcu(&inode->i_rcu, ovl_i_callback);
+}
+
 static void ovl_put_super(struct super_block *sb)
 {
 	struct ovl_fs *ufs = sb->s_fs_info;
 	unsigned i;
 
+	dput(ufs->indexdir);
 	dput(ufs->workdir);
+	ovl_inuse_unlock(ufs->workbasedir);
+	dput(ufs->workbasedir);
+	if (ufs->upper_mnt)
+		ovl_inuse_unlock(ufs->upper_mnt->mnt_root);
 	mntput(ufs->upper_mnt);
 	for (i = 0; i < ufs->numlower; i++)
 		mntput(ufs->lower_mnt[i]);
@@ -228,6 +271,12 @@ static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
 	return err;
 }
 
+/* Will this overlay be forced to mount/remount ro? */
+static bool ovl_force_readonly(struct ovl_fs *ufs)
+{
+	return (!ufs->upper_mnt || !ufs->workdir);
+}
+
 /**
  * ovl_show_options
  *
@@ -249,6 +298,9 @@ static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
 	if (ufs->config.redirect_dir != ovl_redirect_dir_def)
 		seq_printf(m, ",redirect_dir=%s",
 			   ufs->config.redirect_dir ? "on" : "off");
+	if (ufs->config.index != ovl_index_def)
+		seq_printf(m, ",index=%s",
+			   ufs->config.index ? "on" : "off");
 	return 0;
 }
 
@@ -256,19 +308,21 @@ static int ovl_remount(struct super_block *sb, int *flags, char *data)
 {
 	struct ovl_fs *ufs = sb->s_fs_info;
 
-	if (!(*flags & MS_RDONLY) && (!ufs->upper_mnt || !ufs->workdir))
+	if (!(*flags & MS_RDONLY) && ovl_force_readonly(ufs))
 		return -EROFS;
 
 	return 0;
 }
 
 static const struct super_operations ovl_super_operations = {
+	.alloc_inode	= ovl_alloc_inode,
+	.destroy_inode	= ovl_destroy_inode,
+	.drop_inode	= generic_delete_inode,
 	.put_super	= ovl_put_super,
 	.sync_fs	= ovl_sync_fs,
 	.statfs		= ovl_statfs,
 	.show_options	= ovl_show_options,
 	.remount_fs	= ovl_remount,
-	.drop_inode	= generic_delete_inode,
 };
 
 enum {
@@ -278,6 +332,8 @@ enum {
 	OPT_DEFAULT_PERMISSIONS,
 	OPT_REDIRECT_DIR_ON,
 	OPT_REDIRECT_DIR_OFF,
+	OPT_INDEX_ON,
+	OPT_INDEX_OFF,
 	OPT_ERR,
 };
 
@@ -288,6 +344,8 @@ static const match_table_t ovl_tokens = {
 	{OPT_DEFAULT_PERMISSIONS,	"default_permissions"},
 	{OPT_REDIRECT_DIR_ON,		"redirect_dir=on"},
 	{OPT_REDIRECT_DIR_OFF,		"redirect_dir=off"},
+	{OPT_INDEX_ON,			"index=on"},
+	{OPT_INDEX_OFF,			"index=off"},
 	{OPT_ERR,			NULL}
 };
 
@@ -360,6 +418,14 @@ static int ovl_parse_opt(char *opt, struct ovl_config *config)
 			config->redirect_dir = false;
 			break;
 
+		case OPT_INDEX_ON:
+			config->index = true;
+			break;
+
+		case OPT_INDEX_OFF:
+			config->index = false;
+			break;
+
 		default:
 			pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p);
 			return -EINVAL;
@@ -378,23 +444,29 @@ static int ovl_parse_opt(char *opt, struct ovl_config *config)
 }
 
 #define OVL_WORKDIR_NAME "work"
+#define OVL_INDEXDIR_NAME "index"
 
-static struct dentry *ovl_workdir_create(struct vfsmount *mnt,
-					 struct dentry *dentry)
+static struct dentry *ovl_workdir_create(struct super_block *sb,
+					 struct ovl_fs *ufs,
+					 struct dentry *dentry,
+					 const char *name, bool persist)
 {
 	struct inode *dir = dentry->d_inode;
+	struct vfsmount *mnt = ufs->upper_mnt;
 	struct dentry *work;
 	int err;
 	bool retried = false;
+	bool locked = false;
 
 	err = mnt_want_write(mnt);
 	if (err)
-		return ERR_PTR(err);
+		goto out_err;
 
 	inode_lock_nested(dir, I_MUTEX_PARENT);
+	locked = true;
+
 retry:
-	work = lookup_one_len(OVL_WORKDIR_NAME, dentry,
-			      strlen(OVL_WORKDIR_NAME));
+	work = lookup_one_len(name, dentry, strlen(name));
 
 	if (!IS_ERR(work)) {
 		struct iattr attr = {
@@ -407,6 +479,9 @@ retry:
 			if (retried)
 				goto out_dput;
 
+			if (persist)
+				goto out_unlock;
+
 			retried = true;
 			ovl_workdir_cleanup(dir, mnt, work, 0);
 			dput(work);
@@ -446,16 +521,24 @@ retry:
 		inode_unlock(work->d_inode);
 		if (err)
 			goto out_dput;
+	} else {
+		err = PTR_ERR(work);
+		goto out_err;
 	}
 out_unlock:
-	inode_unlock(dir);
 	mnt_drop_write(mnt);
+	if (locked)
+		inode_unlock(dir);
 
 	return work;
 
 out_dput:
 	dput(work);
-	work = ERR_PTR(err);
+out_err:
+	pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
+		ufs->config.workdir, name, -err);
+	sb->s_flags |= MS_RDONLY;
+	work = NULL;
 	goto out_unlock;
 }
 
@@ -555,6 +638,15 @@ static int ovl_lower_dir(const char *name, struct path *path,
 	if (ovl_dentry_remote(path->dentry))
 		*remote = true;
 
+	/*
+	 * The inodes index feature needs to encode and decode file
+	 * handles, so it requires that all layers support them.
+	 */
+	if (ofs->config.index && !ovl_can_decode_fh(path->dentry->d_sb)) {
+		ofs->config.index = false;
+		pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off.\n", name);
+	}
+
 	return 0;
 
 out_put:
@@ -610,7 +702,7 @@ ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
 			size_t size, int flags)
 {
 	struct dentry *workdir = ovl_workdir(dentry);
-	struct inode *realinode = ovl_inode_real(inode, NULL);
+	struct inode *realinode = ovl_inode_real(inode);
 	struct posix_acl *acl = NULL;
 	int err;
 
@@ -652,7 +744,7 @@ ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
 
 	err = ovl_xattr_set(dentry, handler->name, value, size, flags);
 	if (!err)
-		ovl_copyattr(ovl_inode_real(inode, NULL), inode);
+		ovl_copyattr(ovl_inode_real(inode), inode);
 
 	return err;
 
@@ -734,7 +826,6 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	struct path upperpath = { };
 	struct path workpath = { };
 	struct dentry *root_dentry;
-	struct inode *realinode;
 	struct ovl_entry *oe;
 	struct ovl_fs *ufs;
 	struct path *stack = NULL;
@@ -752,8 +843,8 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	if (!ufs)
 		goto out;
 
-	init_waitqueue_head(&ufs->copyup_wq);
 	ufs->config.redirect_dir = ovl_redirect_dir_def;
+	ufs->config.index = ovl_index_def;
 	err = ovl_parse_opt((char *) data, &ufs->config);
 	if (err)
 		goto out_free_config;
@@ -788,9 +879,15 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 		if (err)
 			goto out_put_upperpath;
 
+		err = -EBUSY;
+		if (!ovl_inuse_trylock(upperpath.dentry)) {
+			pr_err("overlayfs: upperdir is in-use by another mount\n");
+			goto out_put_upperpath;
+		}
+
 		err = ovl_mount_dir(ufs->config.workdir, &workpath);
 		if (err)
-			goto out_put_upperpath;
+			goto out_unlock_upperdentry;
 
 		err = -EINVAL;
 		if (upperpath.mnt != workpath.mnt) {
@@ -801,12 +898,20 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 			pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
 			goto out_put_workpath;
 		}
+
+		err = -EBUSY;
+		if (!ovl_inuse_trylock(workpath.dentry)) {
+			pr_err("overlayfs: workdir is in-use by another mount\n");
+			goto out_put_workpath;
+		}
+
+		ufs->workbasedir = workpath.dentry;
 		sb->s_stack_depth = upperpath.mnt->mnt_sb->s_stack_depth;
 	}
 	err = -ENOMEM;
 	lowertmp = kstrdup(ufs->config.lowerdir, GFP_KERNEL);
 	if (!lowertmp)
-		goto out_put_workpath;
+		goto out_unlock_workdentry;
 
 	err = -EINVAL;
 	stacklen = ovl_split_lowerdirs(lowertmp);
@@ -849,20 +954,14 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 			pr_err("overlayfs: failed to clone upperpath\n");
 			goto out_put_lowerpath;
 		}
+
 		/* Don't inherit atime flags */
 		ufs->upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME);
 
 		sb->s_time_gran = ufs->upper_mnt->mnt_sb->s_time_gran;
 
-		ufs->workdir = ovl_workdir_create(ufs->upper_mnt, workpath.dentry);
-		err = PTR_ERR(ufs->workdir);
-		if (IS_ERR(ufs->workdir)) {
-			pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
-				ufs->config.workdir, OVL_WORKDIR_NAME, -err);
-			sb->s_flags |= MS_RDONLY;
-			ufs->workdir = NULL;
-		}
-
+		ufs->workdir = ovl_workdir_create(sb, ufs, workpath.dentry,
+						  OVL_WORKDIR_NAME, false);
 		/*
 		 * Upper should support d_type, else whiteouts are visible.
 		 * Given workdir and upper are on same fs, we can do
@@ -904,6 +1003,13 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 			} else {
 				vfs_removexattr(ufs->workdir, OVL_XATTR_OPAQUE);
 			}
+
+			/* Check if upper/work fs supports file handles */
+			if (ufs->config.index &&
+			    !ovl_can_decode_fh(ufs->workdir->d_sb)) {
+				ufs->config.index = false;
+				pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
+			}
 		}
 	}
 
@@ -941,6 +1047,44 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	else if (ufs->upper_mnt->mnt_sb != ufs->same_sb)
 		ufs->same_sb = NULL;
 
+	if (!(ovl_force_readonly(ufs)) && ufs->config.index) {
+		/* Verify lower root is upper root origin */
+		err = ovl_verify_origin(upperpath.dentry, ufs->lower_mnt[0],
+					stack[0].dentry, false, true);
+		if (err) {
+			pr_err("overlayfs: failed to verify upper root origin\n");
+			goto out_put_lower_mnt;
+		}
+
+		ufs->indexdir = ovl_workdir_create(sb, ufs, workpath.dentry,
+						   OVL_INDEXDIR_NAME, true);
+		err = PTR_ERR(ufs->indexdir);
+		if (IS_ERR(ufs->indexdir))
+			goto out_put_lower_mnt;
+
+		if (ufs->indexdir) {
+			/* Verify upper root is index dir origin */
+			err = ovl_verify_origin(ufs->indexdir, ufs->upper_mnt,
+						upperpath.dentry, true, true);
+			if (err)
+				pr_err("overlayfs: failed to verify index dir origin\n");
+
+			/* Cleanup bad/stale/orphan index entries */
+			if (!err)
+				err = ovl_indexdir_cleanup(ufs->indexdir,
+							   ufs->upper_mnt,
+							   stack, numlower);
+		}
+		if (err || !ufs->indexdir)
+			pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
+		if (err)
+			goto out_put_indexdir;
+	}
+
+	/* Show index=off/on in /proc/mounts for any of the reasons above */
+	if (!ufs->indexdir)
+		ufs->config.index = false;
+
 	if (remote)
 		sb->s_d_op = &ovl_reval_dentry_operations;
 	else
@@ -948,7 +1092,7 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 
 	ufs->creator_cred = cred = prepare_creds();
 	if (!cred)
-		goto out_put_lower_mnt;
+		goto out_put_indexdir;
 
 	/* Never override disk quota limits or use reserved space */
 	cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
@@ -971,12 +1115,13 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	mntput(upperpath.mnt);
 	for (i = 0; i < numlower; i++)
 		mntput(stack[i].mnt);
-	path_put(&workpath);
+	mntput(workpath.mnt);
 	kfree(lowertmp);
 
 	if (upperpath.dentry) {
-		oe->__upperdentry = upperpath.dentry;
-		oe->impure = ovl_is_impuredir(upperpath.dentry);
+		oe->has_upper = true;
+		if (ovl_is_impuredir(upperpath.dentry))
+			ovl_set_flag(OVL_IMPURE, d_inode(root_dentry));
 	}
 	for (i = 0; i < numlower; i++) {
 		oe->lowerstack[i].dentry = stack[i].dentry;
@@ -986,9 +1131,8 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 
 	root_dentry->d_fsdata = oe;
 
-	realinode = d_inode(ovl_dentry_real(root_dentry));
-	ovl_inode_init(d_inode(root_dentry), realinode, !!upperpath.dentry);
-	ovl_copyattr(realinode, d_inode(root_dentry));
+	ovl_inode_init(d_inode(root_dentry), upperpath.dentry,
+		       ovl_dentry_lower(root_dentry));
 
 	sb->s_root = root_dentry;
 
@@ -998,6 +1142,8 @@ out_free_oe:
 	kfree(oe);
 out_put_cred:
 	put_cred(ufs->creator_cred);
+out_put_indexdir:
+	dput(ufs->indexdir);
 out_put_lower_mnt:
 	for (i = 0; i < ufs->numlower; i++)
 		mntput(ufs->lower_mnt[i]);
@@ -1011,8 +1157,12 @@ out_put_lowerpath:
 	kfree(stack);
 out_free_lowertmp:
 	kfree(lowertmp);
+out_unlock_workdentry:
+	ovl_inuse_unlock(workpath.dentry);
 out_put_workpath:
 	path_put(&workpath);
+out_unlock_upperdentry:
+	ovl_inuse_unlock(upperpath.dentry);
 out_put_upperpath:
 	path_put(&upperpath);
 out_free_config:
@@ -1038,14 +1188,43 @@ static struct file_system_type ovl_fs_type = {
 };
 MODULE_ALIAS_FS("overlay");
 
+static void ovl_inode_init_once(void *foo)
+{
+	struct ovl_inode *oi = foo;
+
+	inode_init_once(&oi->vfs_inode);
+}
+
 static int __init ovl_init(void)
 {
-	return register_filesystem(&ovl_fs_type);
+	int err;
+
+	ovl_inode_cachep = kmem_cache_create("ovl_inode",
+					     sizeof(struct ovl_inode), 0,
+					     (SLAB_RECLAIM_ACCOUNT|
+					      SLAB_MEM_SPREAD|SLAB_ACCOUNT),
+					     ovl_inode_init_once);
+	if (ovl_inode_cachep == NULL)
+		return -ENOMEM;
+
+	err = register_filesystem(&ovl_fs_type);
+	if (err)
+		kmem_cache_destroy(ovl_inode_cachep);
+
+	return err;
 }
 
 static void __exit ovl_exit(void)
 {
 	unregister_filesystem(&ovl_fs_type);
+
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
+	kmem_cache_destroy(ovl_inode_cachep);
+
 }
 
 module_init(ovl_init);

fs/overlayfs/util.c

@@ -12,6 +12,10 @@
 #include <linux/slab.h>
 #include <linux/cred.h>
 #include <linux/xattr.h>
+#include <linux/exportfs.h>
+#include <linux/uuid.h>
+#include <linux/namei.h>
+#include <linux/ratelimit.h>
 #include "overlayfs.h"
 #include "ovl_entry.h"
 
@@ -47,6 +51,19 @@ struct super_block *ovl_same_sb(struct super_block *sb)
 	return ofs->same_sb;
 }
 
+bool ovl_can_decode_fh(struct super_block *sb)
+{
+	return (sb->s_export_op && sb->s_export_op->fh_to_dentry &&
+		!uuid_is_null(&sb->s_uuid));
+}
+
+struct dentry *ovl_indexdir(struct super_block *sb)
+{
+	struct ovl_fs *ofs = sb->s_fs_info;
+
+	return ofs->indexdir;
+}
+
 struct ovl_entry *ovl_alloc_entry(unsigned int numlower)
 {
 	size_t size = offsetof(struct ovl_entry, lowerstack[numlower]);
@@ -78,7 +95,7 @@ enum ovl_path_type ovl_path_type(struct dentry *dentry)
 	struct ovl_entry *oe = dentry->d_fsdata;
 	enum ovl_path_type type = 0;
 
-	if (oe->__upperdentry) {
+	if (ovl_dentry_upper(dentry)) {
 		type = __OVL_PATH_UPPER;
 
 		/*
@@ -99,10 +116,9 @@ enum ovl_path_type ovl_path_type(struct dentry *dentry)
 void ovl_path_upper(struct dentry *dentry, struct path *path)
 {
 	struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
-	struct ovl_entry *oe = dentry->d_fsdata;
 
 	path->mnt = ofs->upper_mnt;
-	path->dentry = ovl_upperdentry_dereference(oe);
+	path->dentry = ovl_dentry_upper(dentry);
 }
 
 void ovl_path_lower(struct dentry *dentry, struct path *path)
@@ -125,48 +141,48 @@ enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path)
 }
 
 struct dentry *ovl_dentry_upper(struct dentry *dentry)
+{
+	return ovl_upperdentry_dereference(OVL_I(d_inode(dentry)));
+}
+
+struct dentry *ovl_dentry_lower(struct dentry *dentry)
 {
 	struct ovl_entry *oe = dentry->d_fsdata;
 
-	return ovl_upperdentry_dereference(oe);
+	return oe->numlower ? oe->lowerstack[0].dentry : NULL;
 }
 
-static struct dentry *__ovl_dentry_lower(struct ovl_entry *oe)
+struct dentry *ovl_dentry_real(struct dentry *dentry)
 {
-	return oe->numlower ? oe->lowerstack[0].dentry : NULL;
+	return ovl_dentry_upper(dentry) ?: ovl_dentry_lower(dentry);
 }
 
-struct dentry *ovl_dentry_lower(struct dentry *dentry)
+struct inode *ovl_inode_upper(struct inode *inode)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
+	struct dentry *upperdentry = ovl_upperdentry_dereference(OVL_I(inode));
 
-	return __ovl_dentry_lower(oe);
+	return upperdentry ? d_inode(upperdentry) : NULL;
 }
 
-struct dentry *ovl_dentry_real(struct dentry *dentry)
+struct inode *ovl_inode_lower(struct inode *inode)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
-	struct dentry *realdentry;
-
-	realdentry = ovl_upperdentry_dereference(oe);
-	if (!realdentry)
-		realdentry = __ovl_dentry_lower(oe);
+	return OVL_I(inode)->lower;
+}
 
-	return realdentry;
+struct inode *ovl_inode_real(struct inode *inode)
+{
+	return ovl_inode_upper(inode) ?: ovl_inode_lower(inode);
 }
 
+
 struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
-
-	return oe->cache;
+	return OVL_I(d_inode(dentry))->cache;
 }
 
 void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
-
-	oe->cache = cache;
+	OVL_I(d_inode(dentry))->cache = cache;
 }
 
 bool ovl_dentry_is_opaque(struct dentry *dentry)
@@ -175,23 +191,35 @@ bool ovl_dentry_is_opaque(struct dentry *dentry)
 	return oe->opaque;
 }
 
-bool ovl_dentry_is_impure(struct dentry *dentry)
+bool ovl_dentry_is_whiteout(struct dentry *dentry)
+{
+	return !dentry->d_inode && ovl_dentry_is_opaque(dentry);
+}
+
+void ovl_dentry_set_opaque(struct dentry *dentry)
 {
 	struct ovl_entry *oe = dentry->d_fsdata;
 
-	return oe->impure;
+	oe->opaque = true;
 }
 
-bool ovl_dentry_is_whiteout(struct dentry *dentry)
+/*
+ * For hard links it's possible for ovl_dentry_upper() to return positive, while
+ * there's no actual upper alias for the inode.  Copy up code needs to know
+ * about the existence of the upper alias, so it can't use ovl_dentry_upper().
+ */
+bool ovl_dentry_has_upper_alias(struct dentry *dentry)
 {
-	return !dentry->d_inode && ovl_dentry_is_opaque(dentry);
+	struct ovl_entry *oe = dentry->d_fsdata;
+
+	return oe->has_upper;
 }
 
-void ovl_dentry_set_opaque(struct dentry *dentry)
+void ovl_dentry_set_upper_alias(struct dentry *dentry)
 {
 	struct ovl_entry *oe = dentry->d_fsdata;
 
-	oe->opaque = true;
+	oe->has_upper = true;
 }
 
 bool ovl_redirect_dir(struct super_block *sb)
@@ -203,63 +231,59 @@ bool ovl_redirect_dir(struct super_block *sb)
 
 const char *ovl_dentry_get_redirect(struct dentry *dentry)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
-
-	return oe->redirect;
+	return OVL_I(d_inode(dentry))->redirect;
 }
 
 void ovl_dentry_set_redirect(struct dentry *dentry, const char *redirect)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
+	struct ovl_inode *oi = OVL_I(d_inode(dentry));
 
-	kfree(oe->redirect);
-	oe->redirect = redirect;
+	kfree(oi->redirect);
+	oi->redirect = redirect;
 }
 
-void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry)
+void ovl_inode_init(struct inode *inode, struct dentry *upperdentry,
+		    struct dentry *lowerdentry)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
+	if (upperdentry)
+		OVL_I(inode)->__upperdentry = upperdentry;
+	if (lowerdentry)
+		OVL_I(inode)->lower = d_inode(lowerdentry);
 
-	WARN_ON(!inode_is_locked(upperdentry->d_parent->d_inode));
-	WARN_ON(oe->__upperdentry);
-	/*
-	 * Make sure upperdentry is consistent before making it visible to
-	 * ovl_upperdentry_dereference().
-	 */
-	smp_wmb();
-	oe->__upperdentry = upperdentry;
+	ovl_copyattr(d_inode(upperdentry ?: lowerdentry), inode);
 }
 
-void ovl_inode_init(struct inode *inode, struct inode *realinode, bool is_upper)
+void ovl_inode_update(struct inode *inode, struct dentry *upperdentry)
 {
-	WRITE_ONCE(inode->i_private, (unsigned long) realinode |
-		   (is_upper ? OVL_ISUPPER_MASK : 0));
-}
+	struct inode *upperinode = d_inode(upperdentry);
 
-void ovl_inode_update(struct inode *inode, struct inode *upperinode)
-{
-	WARN_ON(!upperinode);
-	WARN_ON(!inode_unhashed(inode));
-	WRITE_ONCE(inode->i_private,
-		   (unsigned long) upperinode | OVL_ISUPPER_MASK);
-	if (!S_ISDIR(upperinode->i_mode))
+	WARN_ON(OVL_I(inode)->__upperdentry);
+
+	/*
+	 * Make sure upperdentry is consistent before making it visible
+	 */
+	smp_wmb();
+	OVL_I(inode)->__upperdentry = upperdentry;
+	if (!S_ISDIR(upperinode->i_mode) && inode_unhashed(inode)) {
+		inode->i_private = upperinode;
 		__insert_inode_hash(inode, (unsigned long) upperinode);
+	}
 }
 
 void ovl_dentry_version_inc(struct dentry *dentry)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
+	struct inode *inode = d_inode(dentry);
 
-	WARN_ON(!inode_is_locked(dentry->d_inode));
-	oe->version++;
+	WARN_ON(!inode_is_locked(inode));
+	OVL_I(inode)->version++;
 }
 
 u64 ovl_dentry_version_get(struct dentry *dentry)
 {
-	struct ovl_entry *oe = dentry->d_fsdata;
+	struct inode *inode = d_inode(dentry);
 
-	WARN_ON(!inode_is_locked(dentry->d_inode));
-	return oe->version;
+	WARN_ON(!inode_is_locked(inode));
+	return OVL_I(inode)->version;
 }
 
 bool ovl_is_whiteout(struct dentry *dentry)
@@ -276,32 +300,21 @@ struct file *ovl_path_open(struct path *path, int flags)
 
 int ovl_copy_up_start(struct dentry *dentry)
 {
-	struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
-	struct ovl_entry *oe = dentry->d_fsdata;
+	struct ovl_inode *oi = OVL_I(d_inode(dentry));
 	int err;
 
-	spin_lock(&ofs->copyup_wq.lock);
-	err = wait_event_interruptible_locked(ofs->copyup_wq, !oe->copying);
-	if (!err) {
-		if (oe->__upperdentry)
-			err = 1; /* Already copied up */
-		else
-			oe->copying = true;
+	err = mutex_lock_interruptible(&oi->lock);
+	if (!err && ovl_dentry_has_upper_alias(dentry)) {
+		err = 1; /* Already copied up */
+		mutex_unlock(&oi->lock);
 	}
-	spin_unlock(&ofs->copyup_wq.lock);
 
 	return err;
 }
 
 void ovl_copy_up_end(struct dentry *dentry)
 {
-	struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
-	struct ovl_entry *oe = dentry->d_fsdata;
-
-	spin_lock(&ofs->copyup_wq.lock);
-	oe->copying = false;
-	wake_up_locked(&ofs->copyup_wq);
-	spin_unlock(&ofs->copyup_wq.lock);
+	mutex_unlock(&OVL_I(d_inode(dentry))->lock);
 }
 
 bool ovl_check_dir_xattr(struct dentry *dentry, const char *name)
@@ -343,9 +356,8 @@ int ovl_check_setxattr(struct dentry *dentry, struct dentry *upperdentry,
 int ovl_set_impure(struct dentry *dentry, struct dentry *upperdentry)
 {
 	int err;
-	struct ovl_entry *oe = dentry->d_fsdata;
 
-	if (oe->impure)
+	if (ovl_test_flag(OVL_IMPURE, d_inode(dentry)))
 		return 0;
 
 	/*
@@ -355,7 +367,176 @@ int ovl_set_impure(struct dentry *dentry, struct dentry *upperdentry)
 	err = ovl_check_setxattr(dentry, upperdentry, OVL_XATTR_IMPURE,
 				 "y", 1, 0);
 	if (!err)
-		oe->impure = true;
+		ovl_set_flag(OVL_IMPURE, d_inode(dentry));
 
 	return err;
 }
+
+void ovl_set_flag(unsigned long flag, struct inode *inode)
+{
+	set_bit(flag, &OVL_I(inode)->flags);
+}
+
+bool ovl_test_flag(unsigned long flag, struct inode *inode)
+{
+	return test_bit(flag, &OVL_I(inode)->flags);
+}
+
+/**
+ * Caller must hold a reference to inode to prevent it from being freed while
+ * it is marked inuse.
+ */
+bool ovl_inuse_trylock(struct dentry *dentry)
+{
+	struct inode *inode = d_inode(dentry);
+	bool locked = false;
+
+	spin_lock(&inode->i_lock);
+	if (!(inode->i_state & I_OVL_INUSE)) {
+		inode->i_state |= I_OVL_INUSE;
+		locked = true;
+	}
+	spin_unlock(&inode->i_lock);
+
+	return locked;
+}
+
+void ovl_inuse_unlock(struct dentry *dentry)
+{
+	if (dentry) {
+		struct inode *inode = d_inode(dentry);
+
+		spin_lock(&inode->i_lock);
+		WARN_ON(!(inode->i_state & I_OVL_INUSE));
+		inode->i_state &= ~I_OVL_INUSE;
+		spin_unlock(&inode->i_lock);
+	}
+}
+
+/* Called must hold OVL_I(inode)->oi_lock */
+static void ovl_cleanup_index(struct dentry *dentry)
+{
+	struct inode *dir = ovl_indexdir(dentry->d_sb)->d_inode;
+	struct dentry *lowerdentry = ovl_dentry_lower(dentry);
+	struct dentry *upperdentry = ovl_dentry_upper(dentry);
+	struct dentry *index = NULL;
+	struct inode *inode;
+	struct qstr name;
+	int err;
+
+	err = ovl_get_index_name(lowerdentry, &name);
+	if (err)
+		goto fail;
+
+	inode = d_inode(upperdentry);
+	if (inode->i_nlink != 1) {
+		pr_warn_ratelimited("overlayfs: cleanup linked index (%pd2, ino=%lu, nlink=%u)\n",
+				    upperdentry, inode->i_ino, inode->i_nlink);
+		/*
+		 * We either have a bug with persistent union nlink or a lower
+		 * hardlink was added while overlay is mounted. Adding a lower
+		 * hardlink and then unlinking all overlay hardlinks would drop
+		 * overlay nlink to zero before all upper inodes are unlinked.
+		 * As a safety measure, when that situation is detected, set
+		 * the overlay nlink to the index inode nlink minus one for the
+		 * index entry itself.
+		 */
+		set_nlink(d_inode(dentry), inode->i_nlink - 1);
+		ovl_set_nlink_upper(dentry);
+		goto out;
+	}
+
+	inode_lock_nested(dir, I_MUTEX_PARENT);
+	/* TODO: whiteout instead of cleanup to block future open by handle */
+	index = lookup_one_len(name.name, ovl_indexdir(dentry->d_sb), name.len);
+	err = PTR_ERR(index);
+	if (!IS_ERR(index))
+		err = ovl_cleanup(dir, index);
+	inode_unlock(dir);
+	if (err)
+		goto fail;
+
+out:
+	dput(index);
+	return;
+
+fail:
+	pr_err("overlayfs: cleanup index of '%pd2' failed (%i)\n", dentry, err);
+	goto out;
+}
+
+/*
+ * Operations that change overlay inode and upper inode nlink need to be
+ * synchronized with copy up for persistent nlink accounting.
+ */
+int ovl_nlink_start(struct dentry *dentry, bool *locked)
+{
+	struct ovl_inode *oi = OVL_I(d_inode(dentry));
+	const struct cred *old_cred;
+	int err;
+
+	if (!d_inode(dentry) || d_is_dir(dentry))
+		return 0;
+
+	/*
+	 * With inodes index is enabled, we store the union overlay nlink
+	 * in an xattr on the index inode. When whiting out lower hardlinks
+	 * we need to decrement the overlay persistent nlink, but before the
+	 * first copy up, we have no upper index inode to store the xattr.
+	 *
+	 * As a workaround, before whiteout/rename over of a lower hardlink,
+	 * copy up to create the upper index. Creating the upper index will
+	 * initialize the overlay nlink, so it could be dropped if unlink
+	 * or rename succeeds.
+	 *
+	 * TODO: implement metadata only index copy up when called with
+	 *       ovl_copy_up_flags(dentry, O_PATH).
+	 */
+	if (ovl_indexdir(dentry->d_sb) && !ovl_dentry_has_upper_alias(dentry) &&
+	    d_inode(ovl_dentry_lower(dentry))->i_nlink > 1) {
+		err = ovl_copy_up(dentry);
+		if (err)
+			return err;
+	}
+
+	err = mutex_lock_interruptible(&oi->lock);
+	if (err)
+		return err;
+
+	if (!ovl_test_flag(OVL_INDEX, d_inode(dentry)))
+		goto out;
+
+	old_cred = ovl_override_creds(dentry->d_sb);
+	/*
+	 * The overlay inode nlink should be incremented/decremented IFF the
+	 * upper operation succeeds, along with nlink change of upper inode.
+	 * Therefore, before link/unlink/rename, we store the union nlink
+	 * value relative to the upper inode nlink in an upper inode xattr.
+	 */
+	err = ovl_set_nlink_upper(dentry);
+	revert_creds(old_cred);
+
+out:
+	if (err)
+		mutex_unlock(&oi->lock);
+	else
+		*locked = true;
+
+	return err;
+}
+
+void ovl_nlink_end(struct dentry *dentry, bool locked)
+{
+	if (locked) {
+		if (ovl_test_flag(OVL_INDEX, d_inode(dentry)) &&
+		    d_inode(dentry)->i_nlink == 0) {
+			const struct cred *old_cred;
+
+			old_cred = ovl_override_creds(dentry->d_sb);
+			ovl_cleanup_index(dentry);
+			revert_creds(old_cred);
+		}
+
+		mutex_unlock(&OVL_I(d_inode(dentry))->lock);
+	}
+}

include/linux/fs.h

@@ -1955,6 +1955,9 @@ static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
  *			wb stat updates to grab mapping->tree_lock.  See
  *			inode_switch_wb_work_fn() for details.
  *
+ * I_OVL_INUSE		Used by overlayfs to get exclusive ownership on upper
+ *			and work dirs among overlayfs mounts.
+ *
  * Q: What is the difference between I_WILL_FREE and I_FREEING?
  */
 #define I_DIRTY_SYNC		(1 << 0)
@@ -1975,6 +1978,7 @@ static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
 #define __I_DIRTY_TIME_EXPIRED	12
 #define I_DIRTY_TIME_EXPIRED	(1 << __I_DIRTY_TIME_EXPIRED)
 #define I_WB_SWITCH		(1 << 13)
+#define I_OVL_INUSE			(1 << 14)
 
 #define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES)
 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)