vfs: get rid of old '->iterate' directory operation

All users now just use '->iterate_shared()', which only takes the
directory inode lock for reading.

Filesystems that never got convered to shared mode now instead use a
wrapper that drops the lock, re-takes it in write mode, calls the old
function, and then downgrades the lock back to read mode.

This way the VFS layer and other callers no longer need to care about
filesystems that never got converted to the modern era.

The filesystems that use the new wrapper are ceph, coda, exfat, jfs,
ntfs, ocfs2, overlayfs, and vboxsf.

Honestly, several of them look like they really could just iterate their
directories in shared mode and skip the wrapper entirely, but the point
of this change is to not change semantics or fix filesystems that
haven't been fixed in the last 7+ years, but to finally get rid of the
dual iterators.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>

10 files changed, 76 insertions, 39 deletions

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 4a2b39d9a61a..bdcffb04513f 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -2019,9 +2019,10 @@ unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn)
 	}
 }
 
+WRAP_DIR_ITER(ceph_readdir) // FIXME!
 const struct file_operations ceph_dir_fops = {
 	.read = ceph_read_dir,
-	.iterate = ceph_readdir,
+	.iterate_shared = shared_ceph_readdir,
 	.llseek = ceph_dir_llseek,
 	.open = ceph_open,
 	.release = ceph_release,
@@ -2033,7 +2034,7 @@ const struct file_operations ceph_dir_fops = {
 };
 
 const struct file_operations ceph_snapdir_fops = {
-	.iterate = ceph_readdir,
+	.iterate_shared = shared_ceph_readdir,
 	.llseek = ceph_dir_llseek,
 	.open = ceph_open,
 	.release = ceph_release,
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 8450b1bd354b..1b960de2bf39 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -429,21 +429,14 @@ static int coda_readdir(struct file *coda_file, struct dir_context *ctx)
 	cfi = coda_ftoc(coda_file);
 	host_file = cfi->cfi_container;
 
-	if (host_file->f_op->iterate || host_file->f_op->iterate_shared) {
+	if (host_file->f_op->iterate_shared) {
 		struct inode *host_inode = file_inode(host_file);
 		ret = -ENOENT;
 		if (!IS_DEADDIR(host_inode)) {
-			if (host_file->f_op->iterate_shared) {
-				inode_lock_shared(host_inode);
-				ret = host_file->f_op->iterate_shared(host_file, ctx);
-				file_accessed(host_file);
-				inode_unlock_shared(host_inode);
-			} else {
-				inode_lock(host_inode);
-				ret = host_file->f_op->iterate(host_file, ctx);
-				file_accessed(host_file);
-				inode_unlock(host_inode);
-			}
+			inode_lock_shared(host_inode);
+			ret = host_file->f_op->iterate_shared(host_file, ctx);
+			file_accessed(host_file);
+			inode_unlock_shared(host_inode);
 		}
 		return ret;
 	}
@@ -585,10 +578,11 @@ const struct inode_operations coda_dir_inode_operations = {
 	.setattr	= coda_setattr,
 };
 
+WRAP_DIR_ITER(coda_readdir) // FIXME!
 const struct file_operations coda_dir_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-	.iterate	= coda_readdir,
+	.iterate_shared	= shared_coda_readdir,
 	.open		= coda_open,
 	.release	= coda_release,
 	.fsync		= coda_fsync,
diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c
index 598081d0d059..e1586bba6d86 100644
--- a/fs/exfat/dir.c
+++ b/fs/exfat/dir.c
@@ -306,10 +306,11 @@ out:
 	return err;
 }
 
+WRAP_DIR_ITER(exfat_iterate) // FIXME!
 const struct file_operations exfat_dir_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-	.iterate	= exfat_iterate,
+	.iterate_shared	= shared_exfat_iterate,
 	.unlocked_ioctl = exfat_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl = exfat_compat_ioctl,
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c
index 40e624cf7e92..d1dbe47c7975 100644
--- a/fs/exportfs/expfs.c
+++ b/fs/exportfs/expfs.c
@@ -315,7 +315,7 @@ static int get_name(const struct path *path, char *name, struct dentry *child)
 		goto out;
 
 	error = -EINVAL;
-	if (!file->f_op->iterate && !file->f_op->iterate_shared)
+	if (!file->f_op->iterate_shared)
 		goto out_close;
 
 	buffer.sequence = 0;
diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index 9b030297aa64..e98ddb2b1cf2 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -1535,9 +1535,10 @@ const struct inode_operations jfs_dir_inode_operations = {
 #endif
 };
 
+WRAP_DIR_ITER(jfs_readdir) // FIXME!
 const struct file_operations jfs_dir_operations = {
 	.read		= generic_read_dir,
-	.iterate	= jfs_readdir,
+	.iterate_shared	= shared_jfs_readdir,
 	.fsync		= jfs_fsync,
 	.unlocked_ioctl = jfs_ioctl,
 	.compat_ioctl	= compat_ptr_ioctl,
diff --git a/fs/ntfs/dir.c b/fs/ntfs/dir.c
index 518c3a21a556..4596c90e7b7c 100644
--- a/fs/ntfs/dir.c
+++ b/fs/ntfs/dir.c
@@ -1525,10 +1525,11 @@ static int ntfs_dir_fsync(struct file *filp, loff_t start, loff_t end,
 
 #endif /* NTFS_RW */
 
+WRAP_DIR_ITER(ntfs_readdir) // FIXME!
 const struct file_operations ntfs_dir_ops = {
 	.llseek		= generic_file_llseek,	/* Seek inside directory. */
 	.read		= generic_read_dir,	/* Return -EISDIR. */
-	.iterate	= ntfs_readdir,		/* Read directory contents. */
+	.iterate_shared	= shared_ntfs_readdir,	/* Read directory contents. */
 #ifdef NTFS_RW
 	.fsync		= ntfs_dir_fsync,	/* Sync a directory to disk. */
 #endif /* NTFS_RW */
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 91a194596552..bf2c17ea96a0 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -2793,10 +2793,11 @@ const struct file_operations ocfs2_fops = {
 	.remap_file_range = ocfs2_remap_file_range,
 };
 
+WRAP_DIR_ITER(ocfs2_readdir) // FIXME!
 const struct file_operations ocfs2_dops = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-	.iterate	= ocfs2_readdir,
+	.iterate_shared	= shared_ocfs2_readdir,
 	.fsync		= ocfs2_sync_file,
 	.release	= ocfs2_dir_release,
 	.open		= ocfs2_dir_open,
@@ -2842,7 +2843,7 @@ const struct file_operations ocfs2_fops_no_plocks = {
 const struct file_operations ocfs2_dops_no_plocks = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-	.iterate	= ocfs2_readdir,
+	.iterate_shared	= shared_ocfs2_readdir,
 	.fsync		= ocfs2_sync_file,
 	.release	= ocfs2_dir_release,
 	.open		= ocfs2_dir_open,
diff --git a/fs/overlayfs/readdir.c b/fs/overlayfs/readdir.c
index ee5c4736480f..de39e067ae65 100644
--- a/fs/overlayfs/readdir.c
+++ b/fs/overlayfs/readdir.c
@@ -954,10 +954,11 @@ static int ovl_dir_open(struct inode *inode, struct file *file)
 	return 0;
 }
 
+WRAP_DIR_ITER(ovl_iterate) // FIXME!
 const struct file_operations ovl_dir_operations = {
 	.read		= generic_read_dir,
 	.open		= ovl_dir_open,
-	.iterate	= ovl_iterate,
+	.iterate_shared	= shared_ovl_iterate,
 	.llseek		= ovl_dir_llseek,
 	.fsync		= ovl_dir_fsync,
 	.release	= ovl_dir_release,
diff --git a/fs/readdir.c b/fs/readdir.c
index b264ce60114d..c8c46e294431 100644
--- a/fs/readdir.c
+++ b/fs/readdir.c
@@ -25,6 +25,53 @@
 #include <asm/unaligned.h>
 
 /*
+ * Some filesystems were never converted to '->iterate_shared()'
+ * and their directory iterators want the inode lock held for
+ * writing. This wrapper allows for converting from the shared
+ * semantics to the exclusive inode use.
+ */
+int wrap_directory_iterator(struct file *file,
+			    struct dir_context *ctx,
+			    int (*iter)(struct file *, struct dir_context *))
+{
+	struct inode *inode = file_inode(file);
+	int ret;
+
+	/*
+	 * We'd love to have an 'inode_upgrade_trylock()' operation,
+	 * see the comment in mmap_upgrade_trylock() in mm/memory.c.
+	 *
+	 * But considering this is for "filesystems that never got
+	 * converted", it really doesn't matter.
+	 *
+	 * Also note that since we have to return with the lock held
+	 * for reading, we can't use the "killable()" locking here,
+	 * since we do need to get the lock even if we're dying.
+	 *
+	 * We could do the write part killably and then get the read
+	 * lock unconditionally if it mattered, but see above on why
+	 * this does the very simplistic conversion.
+	 */
+	up_read(&inode->i_rwsem);
+	down_write(&inode->i_rwsem);
+
+	/*
+	 * Since we dropped the inode lock, we should do the
+	 * DEADDIR test again. See 'iterate_dir()' below.
+	 *
+	 * Note that we don't need to re-do the f_pos games,
+	 * since the file must be locked wrt f_pos anyway.
+	 */
+	ret = -ENOENT;
+	if (!IS_DEADDIR(inode))
+		ret = iter(file, ctx);
+
+	downgrade_write(&inode->i_rwsem);
+	return ret;
+}
+EXPORT_SYMBOL(wrap_directory_iterator);
+
+/*
  * Note the "unsafe_put_user() semantics: we goto a
  * label for errors.
  */
@@ -40,39 +87,28 @@
 int iterate_dir(struct file *file, struct dir_context *ctx)
 {
 	struct inode *inode = file_inode(file);
-	bool shared = false;
 	int res = -ENOTDIR;
-	if (file->f_op->iterate_shared)
-		shared = true;
-	else if (!file->f_op->iterate)
+
+	if (!file->f_op->iterate_shared)
 		goto out;
 
 	res = security_file_permission(file, MAY_READ);
 	if (res)
 		goto out;
 
-	if (shared)
-		res = down_read_killable(&inode->i_rwsem);
-	else
-		res = down_write_killable(&inode->i_rwsem);
+	res = down_read_killable(&inode->i_rwsem);
 	if (res)
 		goto out;
 
 	res = -ENOENT;
 	if (!IS_DEADDIR(inode)) {
 		ctx->pos = file->f_pos;
-		if (shared)
-			res = file->f_op->iterate_shared(file, ctx);
-		else
-			res = file->f_op->iterate(file, ctx);
+		res = file->f_op->iterate_shared(file, ctx);
 		file->f_pos = ctx->pos;
 		fsnotify_access(file);
 		file_accessed(file);
 	}
-	if (shared)
-		inode_unlock_shared(inode);
-	else
-		inode_unlock(inode);
+	inode_unlock_shared(inode);
 out:
 	return res;
 }
diff --git a/fs/vboxsf/dir.c b/fs/vboxsf/dir.c
index 075f15c43c78..5f1a14d5b927 100644
--- a/fs/vboxsf/dir.c
+++ b/fs/vboxsf/dir.c
@@ -179,9 +179,10 @@ static int vboxsf_dir_iterate(struct file *dir, struct dir_context *ctx)
 	return 0;
 }
 
+WRAP_DIR_ITER(vboxsf_dir_iterate) // FIXME!
 const struct file_operations vboxsf_dir_fops = {
 	.open = vboxsf_dir_open,
-	.iterate = vboxsf_dir_iterate,
+	.iterate_shared = shared_vboxsf_dir_iterate,
 	.release = vboxsf_dir_release,
 	.read = generic_read_dir,
 	.llseek = generic_file_llseek,