What do you mean by "survival of a gene"?

Your first question - use GDCquery - it takes care of loading the expression data:

mrna_query <- GDCquery(project = "TCGA-PRAD",
                       data.category = "Transcriptome Profiling",
                       data.type = "Gene Expression Quantification",
                       workflow.type = "STAR - Counts",
                       experimental.strategy = "RNA-Seq")

GDCdownload(mrna_query, method = "api", files.per.chunk = 100,
            directory = "~/Desktop/TCGA/mRNA/")

mrna_df <- GDCprepare(mrna_query, directory = "~/Desktop/TCGA/mRNA/")

Yes normalisation is required:

y <- edgeR::DGEList(mrna_df)
y <- edgeR::calcNormFactors(y)
logcpm <- edgeR::cpm(y, normalized.lib.sizes = T, log=TRUE)

Follow up on ATpoints comment because he is 100% correct:

Take a look at GEPIA: http://gepia.cancer-pku.cn/index.html to look at survival plots for a single gene in a TCGA dataset

In GEPIA, the two lines in the kaplan-meier plot represent patients stratified into two groups using median gene expression cutoff... each group containing normal and tumor samples...