Why can this linear system (Truss/global stiffness matrix, finding unknown reaction forces and displacements) be solved this way?

Overall System is F=K*u where K (symmetric, z by z, global stiffness matrix), F (z by 1, most elements(some of these are unknown)=0, external forces), u (z by 1, unknown elements=Inf, known elements=0, node displacements)

KK=K; %saves original K, stiffness matrix

cdof=u~=Inf; %so creates logical array for known(constrained degrees of freedom) positioning in u

F(cdof)=u(cdof); %makes corresponding F elements also 0 for some reason

K(cdof,:)=0;

K(:,cdof)=0; %zeros columns and rows corresponding to known displacements

K(cdof,cdof)=eye(sum(cdof)); %changes elements on diagonal that just got changed to 0's, 1's

u=K\F; %no clue why changed K array and F column returns u with unknowns solved

F=KK*u; % understandable why F can be found once u is found