Use bedtools:

$ bedtools intersect -a input.vcf -b genes.gff -header -wa > output.vcf

EDIT:

For (very) large vcf files it might be more efficient to bgzip and tabix index the vcf file, convert your gff to bed and use tabix to query the regions

1. bgzip and index

$ bgzip -c input.vcf > input.vcf.gz
$ tabix input.vcf.gz

2. gff to bed

E.g with BEDOPS:

$ gff2bed < genes.gff > genes.bed

3. Query the regions

$ tabix -R genes.bed -h input.vcf.gz > output.vcf

fin swimmer