To construct the confidence interval you again use a t statistic. In this case the pivotal quantity is (9000 -m)/(300/√25) where m is the true unknown mean. This simplifies to
(9000-m)/60 To construct a two-sided 90% confidence interval we need m to satisfy the condition
t(0.05)< (9000-m)/60 < t(0.95) where t(0.05) is the 5th percentile of the t distribution with n-1 = 24 degrees of freedom and t(0.95) is the 95th percentile for a t distribution with 24 degrees of freedom. Because the t distribution is symmetric about 0 t(0.05)=-t(0.95).
So you take the inequality -t(0.95) < (9000-m)/60 < t(0.95) and rearrange it to be an inequality for m as 9000-60 t(0.95) < m < 9000+60 t(0.95).
The interval [9000-60 t(0.95), 9000+60 t(0.95)] is your answer. From the t distribution with 24 degrees of freedom you will see that t(0.95)=1.711.