This DNS server bypasses the kernel. It has it's own TCP/IP stack that interacts directly with the network adapter. It's designed to service DNS requests at rates of 10 million per second.

Currently it is in "prototype" stage. There is much that almost works, but will still take some effort to finish.

On Debian/Ubuntu, it goes something like this:

$ git clone https://github.com/robertdavidgraham/robdns
$ cd robdns
$ make
$ make regress

This puts the program in the robdns/bin subdirectory. You'll have to manually copy it to something like /usr/local/bin if you want to install it elsewhere on the system.

While Linux is the primary target platform, the code runs well on many other systems. Here's some additional build info:

• Windows w/ Visual Studio: use the VS10 project
• Windows w/ MingGW: just type make

The easiest way to run this server is simply to run it on the command-line with one or more DNS zone-files:

# robdns example.zone

The zone file is assumed to be in standard BIND9 format starting with an SOA record, and containing only records/glue within the zone.

By default, this will use the IP address of the primary network adapter. This causes some difficulties, because incoming packets will be sent both to the normal network stack and to this program. For best results, use a different network address not used by another machine on the local subnet.

# robdns example.con 192.168.1.222

To verify that it's working, use the dig tool from another machine:

$ dig chaos txt version.bind @192.168.1.122

This should return a record with the value of robdns/1. Then, try normal DNS requests, such as:

$ dig ns1.example.com @192.168.1.122

There is a sample example.zone file to test with in the top directory. Or, consider getting a copy of the com.zone file (8-gigabytes) to test with.

This server has no particular features at this time, other than bypassing the kernel.

This tool created by Robert Graham: email: robert_david_graham@yahoo.com twitter: @ErrataRob