I don't understand how he proved it, but here's my solution:
$$f(m-n+f(n))=f(m)+f(n) \implies P(m,n)$$
$$P(0,0) \implies f(f(0))=2f(0)$$
Let $f(0)=c$ giving $f(c)=2c$, so
$$P(m,c)\implies f(m+c)=f(m)+2c $$
$$P(m,0) \implies f(m+c)=f(m)+c $$
So, $c=2c \implies c=0 \implies f(0)=0$
$$P(n,n) \implies f(f(n))=2f(n)$$
Let $f(1)=k$
$$P(1,1) \implies f(k)=2k$$
$$P(m,1) \implies f(m-1+k)=f(m)+k$$
$$P(m-1,k) \implies f(m-1-k+2k)=f(m-1+k)=f(m-1)+2k$$
So, combining the previous 2 equations, we get
$$f(m)+k=f(m-1)+2k \Leftrightarrow f(m)-f(m-1)=k \implies H(m) $$
$$H(2)\implies f(2)-f(1)=k \implies f(2)=2k$$
$$H(3)\implies f(3)-f(2)=k \implies f(3)=k+f(2)=k+2k=3k$$
and so on, by simple induction, and the fact that $f(0)=0$
$$f(x)=kx \text{ } \forall \text{ } x \in \mathbb{N}$$
Substituting in the original equation,
$$km-kn+{k^2}n=km+kn \Leftrightarrow {k^2}n=2kn \Leftrightarrow k^2=2k \Leftrightarrow k^2-2k=0 $$
$$ \Leftrightarrow k(k-2)=0$$
This gives $$(1) \text{ } k=0 \implies f(x)=0 \text{ } \forall \text{ } x \in \mathbb{N}$$
$$(2) \text{ } k=2 \implies f(x)=2x \text{ } \forall \text{ } x \in \mathbb{N}$$
the only 2 solutions $\Box$.
Note that when $0$ doesn't belong to $\mathbb{N}$, you can discard all the steps of relating to $0$ and the proof would still hold true.