Cross-Entropy vs MSE for Classification

Explain why cross-entropy loss is preferred over mean squared error for classification problems, and what goes wrong when you use MSE for classification.