It's in in the general primal-dual algorithm and I don't know why we choose $b \geq 0$ in $P$. I guess it may be related to the RP problem but I am not sure because I don't not have a deep understanding of it. Thanks a lot.
Starting with an LP standard form, which we call the primal problem $P$: \begin{align} \min z = c'\cdot x\\ Ax = b \geq 0\\ x\geq 0 \end{align}