In this case, xtreg y x1 x2 can be referred to both as "regressing y on x1 and x2" and as "regressing y on x1, controlling for x2." The nuance is whether (1) the effect from x2 on y is important for your research question or (2) it's simply a control variable, but what you want to focus on is the effect from x1. If both variables are important, you'll write the first variant (regress ... on ...). If only x1 is important, you'll write the second variant (regress ... on ..., controlling for ...). But in both cases, you'll use the same command.