Bringing the matrix in reduced row echelon form $R$ (the way other answers have shown) is the first step no matter what you do.
As for "reading off" the solutions off $R$, I think there are several techniques. The one I was taught, and which I find the easiest to use, is the following:
- remove all zero rows
In the example this leads to
$\begin{bmatrix}
1 & 0 & -\frac{7}{5} & | & \frac{1}{5} \\
0 & 1 & \frac{3}{5} & | & -\frac{4}{5}
\end{bmatrix}$, so nothing changed here, since there were no zero rows
- then insert any additional zero rows so the matrix is square and the pivots are all on the diagonal
$\begin{bmatrix}
1 & 0 & -\frac{7}{5} & | & \frac{1}{5} \\
0 & 1 & \frac{3}{5} & | & -\frac{4}{5} \\
0 & 0 & 0 & | & 0
\end{bmatrix}$
- now you can read off a special solution $\lambda$ off the right hand side
$\lambda = \begin{pmatrix}\frac{1}{5} \\ -\frac{4}{5} \\ 0\end{pmatrix}$
- replace all zero entries on the diagonal with $-1$.
$\begin{bmatrix}
1 & 0 & -\frac{7}{5} & | & \frac{1}{5} \\
0 & 1 & \frac{3}{5} & | & -\frac{4}{5} \\
0 & 0 & -1 & | & 0
\end{bmatrix}$
- the columns with $-1$ on the diagonal (the free columns) now form the basis of the null space
$N = \langle \begin{pmatrix}-\frac{7}{5} \\ \frac{3}{5} \\ -1\end{pmatrix} \rangle = \{ r\cdot \begin{pmatrix}-\frac{7}{5} \\ \frac{3}{5} \\ -1\end{pmatrix}$ }, for all $r \in \mathbb{R}$.
And then, of course, the complete solution is just $\lambda + N$, that is $L = \{ \begin{pmatrix}\frac{1}{5} \\ -\frac{4}{5} \\ 0\end{pmatrix} + r\cdot \begin{pmatrix}-\frac{7}{5} \\ \frac{3}{5} \\ -1\end{pmatrix} | r \in \mathbb{R} \}$