"the location of a certain gene on a chromosome " is the definition I know for locus. ı just read that a gene is not necessarily present on a locus. how would that be possible?
A locus is not necessarily the position of a gene. It is a position in the genome, whether or not, there is a gene at this position. A locus often refers to the position of a SNP (which may well be outside a coding region) or any other type of genetic marker.
A gene is typically defined as a sequence that is coding for a protein. Most of a eukaryotic genome, such as the human genome is not made of sequences coding for proteins. It is rather made a various regulatory sequences, introns and especially repetitive DNA. So, if you randomly pick a sequence in the human genome, chance are it is not going to be a sequence that gets translated.
Like you said in your question - "locus" means location. A locus may be anywhere on a chromosome - an entire "gene" coding locus, a non-coding locus (junk DNA), a regulatory sequence locus (example - the promoter region of a gene), or just a single nucleotide locus (SNP) etc. A "locus" is the location of interest on the chromosome.
Yes, it is also possible that any genetic "locus" may be absent if it has been lost by a "deletion". A deletion is a type of mutation. They might be a single nucleotide deletion or a very large deletion of thousands of nucleotides. Genetic deletions can happen in several natural ways (a DNA repair error for example), and by other mechanisms, but also through "man made" genetic manipulations.
Locus just means location, as (non-)specifically as in any other use of the word.
Remember the word originates from traditional genetic studies, which didn't know how genes work. Back then, scientists gave some name to "that region in the genome" and they determined the approximate position relative to other known loci through recombination studies. For many of these, we have in retrospect figured out that all of the findings by these older studies were related to a specific gene and now we refer to that gene by the name originally used for the locus. If you read older papers about the white gene in Drosophila, you'll see this usage.
In vertebrates, functionally related genes are often located close to each other (in a cluster) and it's useful to have a name for the whole group. For example, some transcription factor's function might depend on how the genes are arranged in the genomic region. One example is the human beta-globin locus.
Even when a locus contains something like a "gene", it might not be a traditional protein-coding gene but rather a miRNA (or others like tRNA, rRNA) locus. These again sometimes cluster.
The sequence at that location in the genome doesn't need to be functional in any way. For DNA fingerprinting, you could determine the number of VNTR repeats at each VNTR locus.
And finally, like Remi.b mentions, "locus" can very specifically refer to a single position within a gene's sequence.