The indent-line-function variable decide how Emacs will do indentation. I always use indent-relative. If you want to indent relatively in current buffer, execute

 
M-x set-variable RET indent-line-function RET indent-relative
 

The indent-relative always indent to previous line's first nonblank character. In LISP mode, this variable is set as buffer local variable and the value is lisp-indent-line:

 
  (setq-local indent-line-function 'lisp-indent-line)
 

This statement is defined in function lisp-mode-variables. And executed each time the LISP buffer is created, I have to execute set-variable each time which is cumbersome.

The first solution is setq the variable in Emacs init file. But the setq-local will override global setting, so the following configuration in .emacs does not work

 
(setq indent-line-function 'indent-relative)
 

I tried another way like this

 
(add-hook 'lisp-mode-hook
  (lambda ()
    (setq-local indent-line-function 'indent-relative)
  )
)
 

Still don't work.

After a bit try, I figure out the right way to do it

 
(add-hook 'emacs-lisp-mode-hook
          (function (lambda ()
                      (setq-local indent-line-function 'indent-relative)
          )
))
 

I am curious about the difference between lisp-mode-hook and emacs-lisp-mode-hook.

Turns out there are two kinds of LISP in Emacs: the Emacs LISP and general purpose LISP like Common LISP. The file name extension of Emacs LISP is .el and other LISP is .l or .lsp, .lisp.