f2fs: do not recover i_size if it's valid

If i_size is already valid during roll_forward recovery, we should not update
it according to the block alignment.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

fs/f2fs/recovery.c

@@ -425,7 +425,7 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 			continue;
 		}
 
-		if ((start + 1) << PAGE_SHIFT > i_size_read(inode))
+		if (i_size_read(inode) <= (start << PAGE_SHIFT))
 			f2fs_i_size_write(inode, (start + 1) << PAGE_SHIFT);
 
 		/*