I am asked to find all values of constants $m,b \in \mathbb{R}$ such that $f:\mathbb{R}\to \mathbb{R}, x\mapsto mx+b$ is a solution to $f(f(x)) = x$.
So since $f(f(x)) = m(mx+b) + b = m^2x + mb + b$, the equation $$m^2x + mb + b = x$$ must be true. The only way for that to happen, as far as I can see, is that this system of equations must be satisfied: $$ \begin{cases}m^2 = 1 & \\ mb + b = 0 \end{cases} $$ In the case where $m = 1$, we have the bottom equation as $b+b = 0 \iff b = 0$ so $m = 1, b = 0$ is a solution.
In the case where $m = -1$, we have the bottom equation as $-b + b = 0 \iff 0=0$ which meanss that $m=1$, $b \in \mathbb{R}$ is a solution.
Am I missing any other cases/answers? How would this answer be stated in "set-builder notation"?
$ \Big\{m, b \in \mathbb{R} \left| (m = 1\text{ and }b = 0)\text{ or }(m = -1 \text{ and } b \in\mathbb{R} )\Big\}\right. $? Is the extra $b \in\mathbb{R}$ redundant?
In your set-builder notation, you should put $m,b$ into parenthesis, otherwise your set is contained in $\mathbb R$, while you want it to be subset of $\mathbb R^2$.
A correct way, along the lines of your idea, would be $$\{(m,b)\in\mathbb R^2\mid (m = 1, b = 0)\lor (m = -1, b\in\mathbb R)\},$$ but I wouldn't find it aesthetically pleasing or particularly readable. I'd prefer something like this:
$$\{(1,0)\}\cup \{ (m,b)\in\mathbb R^2\mid m = -1\}.$$
Otherwise, everything you wrote is correct.