At least n=200 is needed for a CFA. However, a rule of thumb for both analyses is that you consider 10-15 participants for each item of the questionnaire. This later is more reliable criterion.
There is no fixed rule that applies to all situations as there are many factors that affect the performance of confirmatory factor models such as the type and size of the factor model, the number and quality (reliability) of the indicators (observed variables), the type of estimator used (e.g., maximum likelihood or weighted least squares), the amount and kind of missing data, etc. Even a sample smaller than N = 200 may be fine for certain types of models/situations. If you want to know for sure, the best way to estimate the required sample size is by simulating your model with the expected parameter values using Monte Carlo simulation techniques. I offer a free workshop on sample-size planning using simulations that you can find here:
There are several suggested criteria such as 5 or 10 observations per parameter in you model. You will frequently see a version of this cited in published papers, but note that there is no proven statistical basis for it.