If your covariates file is a file of principal components, you typically include the first 5 or first 10 of any number of the top principal components as covariates. The top principal components usually capture more than 95% of the total variation that is not due to the trait under consideration. If your covariates file contains specific discrete and/or continuous covariates lika age, sex, ethnicity, etc, then you should, in principle, include all.
One can typically determine which principal components to include by running a glm which test for association between the trait of interest and the principal components and then computes p-values for each test.
SNPTEST v2 does have options for inclusion of covariates