Hello Hiroyuki,
The formulas for converting the Euclidean Hessian in a Riemannian Hessian are not really explicit in the book by Absil, Mahony and Sepulchre (2008).
The book still gives the essentials in chapter 5 (from memory) abound second order geometry. There, it is stated how you are supposed to compute the Hessian on a submanifold or on a quotient manifold, which covers most things.
There is a nice, short explicit paper by Absil about computing Hessians for some important cases. I don't think that covers Grassmann, but it should cover the sphere:
"An extrinsic look at the Riemannian Hessian"
For the Grassmannian specifically, where points are represented as orthonormal matrices (which is the standard thing in Manopt), I believe the formulas should be explained to some extent in this paper of mine (and Absil):
"Low-rank matrix completion via preconditioned optimization on the Grassmann manifold"
(skip straight to section 2 --- it's not extremely explicit, and it's not the simplest exposition possible, but it should give good pointers I hope)
Cheers,
Nicolas