I presume you meant that you "don't get any [DE genes from the] FDR corrected ones", rather than not getting any corrected p-values at all. If you're having trouble finding them, they're in the adj.P.Val column produced by topTable.

... and in any scenario where you have a moderate-to-large number of tests. For example, if you have 1000 genes that are not DE, then you can expect that 50 of them will have a p-value below 0.05, just by chance. The FDR adjustment protects you against this multiplicity effect and, as b.nota says, is mandatory for limma (and really, any genomics analysis with thousands of features). If you don't get any DE genes from the adjusted p-values, then so be it; your system either doesn't have any DE genes, or your study doesn't have enough power to detect them. Forgoing the multiplicity correction isn't going to change that.

If you want, you can relax the FDR threshold to, say, 20%, to identify more putative DE genes. This will give you a larger set of genes, albeit with a greater proportion of false positives. However, at least you're upfront about the fact that your discoveries are more likely to be false. This is preferable to trying to sneak in "DE" genes at the typical 5% threshold using unadjusted p-values, which is rather disingenuous to me. Say you give your collaborators a list of DE genes and tell them that the error rate is 5%, but it actually ends up being 50-100% after they waste a fortune on follow-up experiments - imagine how happy they would (not) be.