As with any linear problem the answer is "take something which satisfies the RHS and add something from the kernel".
So here: Find a function of (x,y) which satisfies [itex]\nabla^2 f = -2[/itex] and add solutions of Laplace's equation to satisfy the boundary condition. This is easier if your particular solution satisfies as much of the boundary condition as possible. Here I think you have less work if your function vanishes on [itex]x = 0[/itex] and [itex]x = 1[/itex].
Here you have [itex]u(x,0) = u(x,1) = u(1,-x) = -\sinh \pi \sin(\pi x)[/itex].
Don't forget that you can use [itex]\cosh (kx)[/itex] and [itex]\sinh(kx)[/itex] in place of [itex]e^{\pm kx}[/itex] and that [itex]\sinh(k(1 - x))[/itex] is a linear combination of these.