It's a linear equation (supposed that [itex]a, \quad b, \quad c[/itex] are functions of the independent variable [itex]x[/itex] but not of the unknown function [itex]y[/itex]), which makes the task to solve it easier.
First of all, suppose you have two solutions [itex]y_1(x)[/itex] and [itex]y_2(x)[/itex]. Now consider [itex]y(x)=y_1(x)-y_2(x)[/itex]. Now, because differentiation wrt. [itex]x[/itex] is a linear operation we have
[tex]a y''+b y' + c y = (a y_1''+by_1'+c y_1)-(a y_2''+b y_2' +c y_2)=f-f=0.[/tex]
This means that the difference of two solutions of the inhomogeneous equations is always a solution of the homogeneous equations.
In turn we can conclude that any solution of the inhomogeneous equation is given as the sum of one particular solution of the inhomogeneous equation [itex]y_p(x)[/itex] and the general solution of the homogeneous equation [itex]y_h(x)[/itex].
Further it's clear that with any set of solutions of the homogeneous equation also any linear combination of such solutions is again a solution of the homogeneous equation. The solutions of the homogeneous equation thus build a linear subspace in the vector space of twice differentiable functions. One can prove that this subspace is two-dimensional (look for "Wronskian Determinant" in your textbook or online), i.e., you need to find two linearly independent solutions of the homogeneous equation [itex]y_{h1}(x)[/itex] and [itex]y_{h2}(x)[/itex]. Linearly independent means simply that [itex]y_{h1}(x)/y_{h2}(x) \neq \text{const}[/itex]. Then the general solution of the inhomogeneous system reads
[tex]y(t)=C_1 y_{h1}(x) + C_2 y_{h2}(x)+y_p(x).[/tex]
Here, [itex]C_1[/itex] and [itex]C_2[/itex] are arbitrary constants.
A solution can be uniquely determined by imposing, e.g., initial conditions,
[tex]y(t_0)=y_0, \quad y'(t_0)=y_0'.[/tex]