I am recently read the differential form book of do carmo and found the following

Here I can not understand what is $(dx_i)_p$ here?Is it the derivative map of $x_i$.
And I also can not understand that how did author calculate $$(dx_i)_p(e_j)=\frac{\partial x_i}{\partial x_j}.$$
And how did become it the dual basis of {$(e_i)_p$}.
Can someone explain more clearly please
It's usual to regard what he's written, as the definition (see e.g. Introduction to Smooth Manifolds, John M Lee). I believe do Carmo's trying to connect up with the total differential/derivative as defined in calculus courses (e.g. Thomas & Finney) where suitably interpreted expressions such as $df = f_1dx_1 + \cdots+f_ndx_n$, the $f_i$'s are the partials of $f$, occur and the $dx_i$ which represent the increments in the $x_i$ variables are taken to be "independent" of one another. Writing $(df)_p$ as the $1\times n$ matrix $[f_1(p) \ \dots f_n(p)]$, we see that $(df)_p$ can be regarded as a linear functional acting on $\mathbb{R}^n$ and the $(dx_i)_p = [\frac{\partial{x_i}}{\partial x_1} \ \cdots \ \frac{\partial{x_i}}{\partial x_n}] = [0 \ \cdots 1\ \cdots 0]$ can be taken to be the basis vectors of the dual of $\mathbb{R}^n$. It should now be clear that $(dx_i)_p(e_j) = \frac{\partial{x_i}}{\partial x_j} = \delta_{ij}$ (Kronecker delta).