This is to reconcile Albert's solution with mine.
[sp]I defined a primitive solution to be one in which the coefficients $a$, $b$, $c$ have no common factor (other than $\pm1$). In Albert's family of equations $2bx^2 + bx = 0$, the only primitive solutions are those where $b = \pm1.$ In particular, if $b = -1$ then the equation becomes $(-2)x^2 + (-1)x + 0 = 0$, which occurs in my family of primitive solutions by taking $m= -2$ and $n=1$.
That example raises something that I overlooked in my solution. My formula for the common difference $k$ in the AP is $k = m^2 + 3mn + 3n^2$. That expression is positive definite, so my formula only covers equations whose coefficients form an AP in which the common difference is positive. But if $ax^2 + bx + c = 0$ has coefficients forming an AP with negative common difference then by taking its negative you get the equation $(-a)x^2 + (-b)x + (-c) = 0$. That is obviously just a different way of writing the same equation, but this time the coefficients form an AP with positive common difference.
So I need to reformulate the definition of a primitive solution to be one in which $a$, $b$, $c$ have no common factor (other than $\pm1$) and the common difference $b-a$ is positive. Then the formula $$a = mn,$$ $$b = m^2 + 4mn + 3n^2,$$ $$c = 2m^2 + 7mn + 6n^2,$$ where $m$ and $n$ are nonzero co-prime integers, $m$ is not a multiple of $3$, and $n>0$, gives what the OP wanted, namely a parametric form for the primitive solutions. The general solution is then obtained by multiplying all the coefficients in a primitive solution by some nonzero integer $d$.[/sp]