Feature selection methodology

I have three sets of data, a training set, validation set, and a test set. The validation set is used to optimize the parameters of the classifier.

Before I got the validation set I was using the performance on leave-one-out cross validation on the training set to select my features.

My question is, is the validation set only used to optimize the parameters of the classifier, like C in a SVM, or can I also use it to select features? It makes sense to me that I could use it but I have never specifically seen someone select their features at that stage.