prediction using random forest model returned NA

[amro1]

• when predicting the test sets in cross validation splits using randomforest function, it returns NA for all values in the response variable .
  here is the code I used for building the model:
  cv_model_tunerf <- cv_tune %>%
  mutate(model = map2(.x = train, .y = mtry, ~randomForest(formula = main_cat ~ . - sample_no., data = .x, ntree = 500, mtry = .y, importance = TRUE, na.action=na.roughfix, seed = 12345)))
  it works well.
• here is the code I used for prediction
  cv_pred_rf <- cv_models_tunerf %>%
  mutate(validate_predicted = map2(.x = model, .y = validate, ~predict(.x, .y)))
  it returned NA for all predicted values.
  note: the dataset I am working on contains many misssing values.
  thank you in advance.