Merge branch 'xfs-4.7-optimise-inline-symlinks' into for-next

fs/namei.c

@@ -4515,7 +4515,6 @@ int readlink_copy(char __user *buffer, int buflen, const char *link)
 out:
 	return len;
 }
-EXPORT_SYMBOL(readlink_copy);
 
 /*
  * A helper for ->readlink().  This should be used *ONLY* for symlinks that

fs/xfs/libxfs/xfs_dir2_sf.c

@@ -257,15 +257,12 @@ xfs_dir2_block_to_sf(
 	 *
 	 * Convert the inode to local format and copy the data in.
 	 */
-	dp->i_df.if_flags &= ~XFS_IFEXTENTS;
-	dp->i_df.if_flags |= XFS_IFINLINE;
-	dp->i_d.di_format = XFS_DINODE_FMT_LOCAL;
 	ASSERT(dp->i_df.if_bytes == 0);
-	xfs_idata_realloc(dp, size, XFS_DATA_FORK);
+	xfs_init_local_fork(dp, XFS_DATA_FORK, dst, size);
+	dp->i_d.di_format = XFS_DINODE_FMT_LOCAL;
+	dp->i_d.di_size = size;
 
 	logflags |= XFS_ILOG_DDATA;
-	memcpy(dp->i_df.if_u1.if_data, dst, size);
-	dp->i_d.di_size = size;
 	xfs_dir2_sf_check(args);
 out:
 	xfs_trans_log_inode(args->trans, dp, logflags);

fs/xfs/libxfs/xfs_inode_fork.c

@@ -231,6 +231,48 @@ xfs_iformat_fork(
 	return error;
 }
 
+void
+xfs_init_local_fork(
+	struct xfs_inode	*ip,
+	int			whichfork,
+	const void		*data,
+	int			size)
+{
+	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
+	int			mem_size = size, real_size = 0;
+	bool			zero_terminate;
+
+	/*
+	 * If we are using the local fork to store a symlink body we need to
+	 * zero-terminate it so that we can pass it back to the VFS directly.
+	 * Overallocate the in-memory fork by one for that and add a zero
+	 * to terminate it below.
+	 */
+	zero_terminate = S_ISLNK(VFS_I(ip)->i_mode);
+	if (zero_terminate)
+		mem_size++;
+
+	if (size == 0)
+		ifp->if_u1.if_data = NULL;
+	else if (mem_size <= sizeof(ifp->if_u2.if_inline_data))
+		ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
+	else {
+		real_size = roundup(mem_size, 4);
+		ifp->if_u1.if_data = kmem_alloc(real_size, KM_SLEEP | KM_NOFS);
+	}
+
+	if (size) {
+		memcpy(ifp->if_u1.if_data, data, size);
+		if (zero_terminate)
+			ifp->if_u1.if_data[size] = '\0';
+	}
+
+	ifp->if_bytes = size;
+	ifp->if_real_bytes = real_size;
+	ifp->if_flags &= ~(XFS_IFEXTENTS | XFS_IFBROOT);
+	ifp->if_flags |= XFS_IFINLINE;
+}
+
 /*
  * The file is in-lined in the on-disk inode.
  * If it fits into if_inline_data, then copy
@@ -248,8 +290,6 @@ xfs_iformat_local(
 	int		whichfork,
 	int		size)
 {
-	xfs_ifork_t	*ifp;
-	int		real_size;
 
 	/*
 	 * If the size is unreasonable, then something
@@ -265,22 +305,8 @@ xfs_iformat_local(
 				     ip->i_mount, dip);
 		return -EFSCORRUPTED;
 	}
-	ifp = XFS_IFORK_PTR(ip, whichfork);
-	real_size = 0;
-	if (size == 0)
-		ifp->if_u1.if_data = NULL;
-	else if (size <= sizeof(ifp->if_u2.if_inline_data))
-		ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
-	else {
-		real_size = roundup(size, 4);
-		ifp->if_u1.if_data = kmem_alloc(real_size, KM_SLEEP | KM_NOFS);
-	}
-	ifp->if_bytes = size;
-	ifp->if_real_bytes = real_size;
-	if (size)
-		memcpy(ifp->if_u1.if_data, XFS_DFORK_PTR(dip, whichfork), size);
-	ifp->if_flags &= ~XFS_IFEXTENTS;
-	ifp->if_flags |= XFS_IFINLINE;
+
+	xfs_init_local_fork(ip, whichfork, XFS_DFORK_PTR(dip, whichfork), size);
 	return 0;
 }
 

fs/xfs/libxfs/xfs_inode_fork.h

@@ -134,6 +134,7 @@ void		xfs_iroot_realloc(struct xfs_inode *, int, int);
 int		xfs_iread_extents(struct xfs_trans *, struct xfs_inode *, int);
 int		xfs_iextents_copy(struct xfs_inode *, struct xfs_bmbt_rec *,
 				  int);
+void		xfs_init_local_fork(struct xfs_inode *, int, const void *, int);
 
 struct xfs_bmbt_rec_host *
 		xfs_iext_get_ext(struct xfs_ifork *, xfs_extnum_t);

fs/xfs/xfs_inode.c

@@ -2840,6 +2840,7 @@ xfs_rename_alloc_whiteout(
 	 * and flag it as linkable.
 	 */
 	drop_nlink(VFS_I(tmpfile));
+	xfs_setup_iops(tmpfile);
 	xfs_finish_inode_setup(tmpfile);
 	VFS_I(tmpfile)->i_state |= I_LINKABLE;
 

fs/xfs/xfs_inode.h

@@ -440,14 +440,16 @@ loff_t	__xfs_seek_hole_data(struct inode *inode, loff_t start,
 
 
 /* from xfs_iops.c */
+extern void xfs_setup_inode(struct xfs_inode *ip);
+extern void xfs_setup_iops(struct xfs_inode *ip);
+
 /*
  * When setting up a newly allocated inode, we need to call
  * xfs_finish_inode_setup() once the inode is fully instantiated at
  * the VFS level to prevent the rest of the world seeing the inode
  * before we've completed instantiation. Otherwise we can do it
  * the moment the inode lookup is complete.
  */
-extern void xfs_setup_inode(struct xfs_inode *ip);
 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
 {
 	xfs_iflags_clear(ip, XFS_INEW);
@@ -458,6 +460,7 @@ static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
 {
 	xfs_setup_inode(ip);
+	xfs_setup_iops(ip);
 	xfs_finish_inode_setup(ip);
 }
 

fs/xfs/xfs_inode_item.c

@@ -210,7 +210,7 @@ xfs_inode_item_format_data_fork(
 			 */
 			data_bytes = roundup(ip->i_df.if_bytes, 4);
 			ASSERT(ip->i_df.if_real_bytes == 0 ||
-			       ip->i_df.if_real_bytes == data_bytes);
+			       ip->i_df.if_real_bytes >= data_bytes);
 			ASSERT(ip->i_df.if_u1.if_data != NULL);
 			ASSERT(ip->i_d.di_size > 0);
 			xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL,
@@ -305,7 +305,7 @@ xfs_inode_item_format_attr_fork(
 			 */
 			data_bytes = roundup(ip->i_afp->if_bytes, 4);
 			ASSERT(ip->i_afp->if_real_bytes == 0 ||
-			       ip->i_afp->if_real_bytes == data_bytes);
+			       ip->i_afp->if_real_bytes >= data_bytes);
 			ASSERT(ip->i_afp->if_u1.if_data != NULL);
 			xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL,
 					ip->i_afp->if_u1.if_data,

fs/xfs/xfs_ioctl.c

@@ -277,7 +277,6 @@ xfs_readlink_by_handle(
 {
 	struct dentry		*dentry;
 	__u32			olen;
-	void			*link;
 	int			error;
 
 	if (!capable(CAP_SYS_ADMIN))
@@ -288,7 +287,7 @@ xfs_readlink_by_handle(
 		return PTR_ERR(dentry);
 
 	/* Restrict this handle operation to symlinks only. */
-	if (!d_is_symlink(dentry)) {
+	if (!d_inode(dentry)->i_op->readlink) {
 		error = -EINVAL;
 		goto out_dput;
 	}
@@ -298,21 +297,8 @@ xfs_readlink_by_handle(
 		goto out_dput;
 	}
 
-	link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
-	if (!link) {
-		error = -ENOMEM;
-		goto out_dput;
-	}
-
-	error = xfs_readlink(XFS_I(d_inode(dentry)), link);
-	if (error)
-		goto out_kfree;
-	error = readlink_copy(hreq->ohandle, olen, link);
-	if (error)
-		goto out_kfree;
+	error = d_inode(dentry)->i_op->readlink(dentry, hreq->ohandle, olen);
 
- out_kfree:
-	kfree(link);
  out_dput:
 	dput(dentry);
 	return error;

fs/xfs/xfs_iops.c

@@ -181,6 +181,8 @@ xfs_generic_create(
 	}
 #endif
 
+	xfs_setup_iops(ip);
+
 	if (tmpfile)
 		d_tmpfile(dentry, inode);
 	else
@@ -368,6 +370,8 @@ xfs_vn_symlink(
 	if (unlikely(error))
 		goto out_cleanup_inode;
 
+	xfs_setup_iops(cip);
+
 	d_instantiate(dentry, inode);
 	xfs_finish_inode_setup(cip);
 	return 0;
@@ -442,6 +446,16 @@ xfs_vn_get_link(
 	return ERR_PTR(error);
 }
 
+STATIC const char *
+xfs_vn_get_link_inline(
+	struct dentry		*dentry,
+	struct inode		*inode,
+	struct delayed_call	*done)
+{
+	ASSERT(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE);
+	return XFS_I(inode)->i_df.if_u1.if_data;
+}
+
 STATIC int
 xfs_vn_getattr(
 	struct vfsmount		*mnt,
@@ -1160,6 +1174,18 @@ static const struct inode_operations xfs_symlink_inode_operations = {
 	.update_time		= xfs_vn_update_time,
 };
 
+static const struct inode_operations xfs_inline_symlink_inode_operations = {
+	.readlink		= generic_readlink,
+	.get_link		= xfs_vn_get_link_inline,
+	.getattr		= xfs_vn_getattr,
+	.setattr		= xfs_vn_setattr,
+	.setxattr		= generic_setxattr,
+	.getxattr		= generic_getxattr,
+	.removexattr		= generic_removexattr,
+	.listxattr		= xfs_vn_listxattr,
+	.update_time		= xfs_vn_update_time,
+};
+
 STATIC void
 xfs_diflags_to_iflags(
 	struct inode		*inode,
@@ -1186,7 +1212,7 @@ xfs_diflags_to_iflags(
 }
 
 /*
- * Initialize the Linux inode and set up the operation vectors.
+ * Initialize the Linux inode.
  *
  * When reading existing inodes from disk this is called directly from xfs_iget,
  * when creating a new inode it is called from xfs_ialloc after setting up the
@@ -1225,32 +1251,12 @@ xfs_setup_inode(
 	i_size_write(inode, ip->i_d.di_size);
 	xfs_diflags_to_iflags(inode, ip);
 
-	ip->d_ops = ip->i_mount->m_nondir_inode_ops;
-	lockdep_set_class(&ip->i_lock.mr_lock, &xfs_nondir_ilock_class);
-	switch (inode->i_mode & S_IFMT) {
-	case S_IFREG:
-		inode->i_op = &xfs_inode_operations;
-		inode->i_fop = &xfs_file_operations;
-		inode->i_mapping->a_ops = &xfs_address_space_operations;
-		break;
-	case S_IFDIR:
+	if (S_ISDIR(inode->i_mode)) {
 		lockdep_set_class(&ip->i_lock.mr_lock, &xfs_dir_ilock_class);
-		if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
-			inode->i_op = &xfs_dir_ci_inode_operations;
-		else
-			inode->i_op = &xfs_dir_inode_operations;
-		inode->i_fop = &xfs_dir_file_operations;
 		ip->d_ops = ip->i_mount->m_dir_inode_ops;
-		break;
-	case S_IFLNK:
-		inode->i_op = &xfs_symlink_inode_operations;
-		if (!(ip->i_df.if_flags & XFS_IFINLINE))
-			inode->i_mapping->a_ops = &xfs_address_space_operations;
-		break;
-	default:
-		inode->i_op = &xfs_inode_operations;
-		init_special_inode(inode, inode->i_mode, inode->i_rdev);
-		break;
+	} else {
+		ip->d_ops = ip->i_mount->m_nondir_inode_ops;
+		lockdep_set_class(&ip->i_lock.mr_lock, &xfs_nondir_ilock_class);
 	}
 
 	/*
@@ -1270,3 +1276,35 @@ xfs_setup_inode(
 		cache_no_acl(inode);
 	}
 }
+
+void
+xfs_setup_iops(
+	struct xfs_inode	*ip)
+{
+	struct inode		*inode = &ip->i_vnode;
+
+	switch (inode->i_mode & S_IFMT) {
+	case S_IFREG:
+		inode->i_op = &xfs_inode_operations;
+		inode->i_fop = &xfs_file_operations;
+		inode->i_mapping->a_ops = &xfs_address_space_operations;
+		break;
+	case S_IFDIR:
+		if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
+			inode->i_op = &xfs_dir_ci_inode_operations;
+		else
+			inode->i_op = &xfs_dir_inode_operations;
+		inode->i_fop = &xfs_dir_file_operations;
+		break;
+	case S_IFLNK:
+		if (ip->i_df.if_flags & XFS_IFINLINE)
+			inode->i_op = &xfs_inline_symlink_inode_operations;
+		else
+			inode->i_op = &xfs_symlink_inode_operations;
+		break;
+	default:
+		inode->i_op = &xfs_inode_operations;
+		init_special_inode(inode, inode->i_mode, inode->i_rdev);
+		break;
+	}
+}

fs/xfs/xfs_symlink.c

@@ -131,6 +131,8 @@ xfs_readlink(
 
 	trace_xfs_readlink(ip);
 
+	ASSERT(!(ip->i_df.if_flags & XFS_IFINLINE));
+
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
 
@@ -150,12 +152,7 @@ xfs_readlink(
 	}
 
 
-	if (ip->i_df.if_flags & XFS_IFINLINE) {
-		memcpy(link, ip->i_df.if_u1.if_data, pathlen);
-		link[pathlen] = '\0';
-	} else {
-		error = xfs_readlink_bmap(ip, link);
-	}
+	error = xfs_readlink_bmap(ip, link);
 
  out:
 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
@@ -303,19 +300,11 @@ xfs_symlink(
 	 * If the symlink will fit into the inode, write it inline.
 	 */
 	if (pathlen <= XFS_IFORK_DSIZE(ip)) {
-		xfs_idata_realloc(ip, pathlen, XFS_DATA_FORK);
-		memcpy(ip->i_df.if_u1.if_data, target_path, pathlen);
-		ip->i_d.di_size = pathlen;
-
-		/*
-		 * The inode was initially created in extent format.
-		 */
-		ip->i_df.if_flags &= ~(XFS_IFEXTENTS | XFS_IFBROOT);
-		ip->i_df.if_flags |= XFS_IFINLINE;
+		xfs_init_local_fork(ip, XFS_DATA_FORK, target_path, pathlen);
 
+		ip->i_d.di_size = pathlen;
 		ip->i_d.di_format = XFS_DINODE_FMT_LOCAL;
 		xfs_trans_log_inode(tp, ip, XFS_ILOG_DDATA | XFS_ILOG_CORE);
-
 	} else {
 		int	offset;