12 KiB
4 - Baby's first DNS server
Haven gotten this far, we're ready to make our first attempt at writing an actual server. Real DNS servers come in two different varieties:
- Authoritative Server - A DNS server hosting one or more "zones". For instance, the authoritative servers for the zone google.com are ns1.google.com, ns2.google.com, ns3.google.com and ns4.google.com.
- Caching Server - A DNS server that services DNS lookups by first checking its cache to see if it already knows of the record being requested, and if not performing a recursive lookup to figure it out. This includes the DNS server that is likely running on your home router as well as the DNS server that your ISP assigns to you through DHCP, and Google's public DNS servers 8.8.8.8 and 8.8.4.4.
Strictly speaking, there's nothing to stop a server from doing both things, but
in practice these two roles are typically mutually exclusive. This also explains
the significance of the flags RD (Recursion Desired) and RA (Recursion
Available) in the packet header -- a stub resolver querying a caching server
will set the RD flag, and since the server allows such queries it will
perform the lookup and send a reply with the RA flag set. This won't work for
an Authoritative Server which will only reply to queries relating to the zones
hosted, and as such will send an error response to any queries with the RD
flag set.
Don't take my word for it, though! Let's verify that this is the case. First
off, let's use 8.8.8.8 for looking up yahoo.com:
# dig @8.8.8.8 yahoo.com
; <<>> DiG 9.10.3-P4-Ubuntu <<>> +recurse @8.8.8.8 yahoo.com
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 53231
;; flags: qr rd ra; QUERY: 1, ANSWER: 3, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 512
;; QUESTION SECTION:
;yahoo.com. IN A
;; ANSWER SECTION:
yahoo.com. 1051 IN A 98.138.253.109
yahoo.com. 1051 IN A 98.139.183.24
yahoo.com. 1051 IN A 206.190.36.45
;; Query time: 1 msec
;; SERVER: 8.8.8.8#53(8.8.8.8)
;; WHEN: Fri Jul 08 11:43:55 CEST 2016
;; MSG SIZE rcvd: 86
This works as expected. Now let's try sending the same query to one of the servers hosting the google.com zone:
# dig @ns1.google.com yahoo.com
; <<>> DiG 9.10.3-P4-Ubuntu <<>> +recurse @ns1.google.com yahoo.com
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: REFUSED, id: 12034
;; flags: qr rd; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 0
;; WARNING: recursion requested but not available
;; QUESTION SECTION:
;yahoo.com. IN A
;; Query time: 10 msec
;; SERVER: 216.239.32.10#53(216.239.32.10)
;; WHEN: Fri Jul 08 11:44:07 CEST 2016
;; MSG SIZE rcvd: 27
Notice how the status of the response says REFUSED! dig also warns us that
while the RD flag was set in the query, the server didn't set the RA flag in the
response. We can still use the same server for google.com, however:
dig @ns1.google.com google.com <<<
; <<>> DiG 9.10.3-P4-Ubuntu <<>> +recurse @ns1.google.com google.com
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 28058
;; flags: qr aa rd; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0
;; WARNING: recursion requested but not available
;; QUESTION SECTION:
;google.com. IN A
;; ANSWER SECTION:
google.com. 300 IN A 216.58.211.142
;; Query time: 10 msec
;; SERVER: 216.239.32.10#53(216.239.32.10)
;; WHEN: Fri Jul 08 11:46:27 CEST 2016
;; MSG SIZE rcvd: 44
No error this time -- however, dig still warns us that recursion is
unavailable. We can explicitly unset it using +norecurse which gets rid of
the warning:
# dig +norecurse @ns1.google.com google.com
; <<>> DiG 9.10.3-P4-Ubuntu <<>> +norecurse @ns1.google.com google.com
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 15850
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0
;; QUESTION SECTION:
;google.com. IN A
;; ANSWER SECTION:
google.com. 300 IN A 216.58.211.142
;; Query time: 10 msec
;; SERVER: 216.239.32.10#53(216.239.32.10)
;; WHEN: Fri Jul 08 11:47:52 CEST 2016
;; MSG SIZE rcvd: 44
This final query is the type of query that we'd expect to see a caching server send as part of recursively resolving the name.
For our first foray into writing our own server, we'll do something even simpler by implementing a server that simply forwards queries to another caching server, i.e. a "DNS proxy server". Having already done most of the hard work, it's a rather quick effort!
Separating lookup into a separate function
We'll start out by doing some quick refactoring, moving our lookup code into
a separate function. This is for the most part the same code as we had in our
main function in the previous chapter, with the only change being that we
handle errors gracefully using try!.
fn lookup(qname: &str, qtype: QueryType, server: (&str, u16)) -> Result<DnsPacket> {
let socket = try!(UdpSocket::bind(("0.0.0.0", 43210)));
let mut packet = DnsPacket::new();
packet.header.id = 6666;
packet.header.questions = 1;
packet.header.recursion_desired = true;
packet.questions.push(DnsQuestion::new(qname.to_string(), qtype));
let mut req_buffer = BytePacketBuffer::new();
packet.write(&mut req_buffer).unwrap();
try!(socket.send_to(&req_buffer.buf[0..req_buffer.pos], server));
let mut res_buffer = BytePacketBuffer::new();
socket.recv_from(&mut res_buffer.buf).unwrap();
DnsPacket::from_buffer(&mut res_buffer)
}
Implementing our first server
Now we'll write our server code. First, we need get some things in order.
fn main() {
// Forward queries to Google's public DNS
let server = ("8.8.8.8", 53);
// Bind an UDP socket on port 2053
let socket = UdpSocket::bind(("0.0.0.0", 2053)).unwrap();
// For now, queries are handled sequentially, so an infinite loop for servicing
// requests is initiated.
loop {
// With a socket ready, we can go ahead and read a packet. This will
// block until one is received.
let mut req_buffer = BytePacketBuffer::new();
let (_, src) = match socket.recv_from(&mut req_buffer.buf) {
Ok(x) => x,
Err(e) => {
println!("Failed to read from UDP socket: {:?}", e);
continue;
}
};
// Here we use match to safely unwrap the `Result`. If everything's as expected,
// the raw bytes are simply returned, and if not it'll abort by restarting the
// loop and waiting for the next request. The `recv_from` function will write the
// data into the provided buffer, and return the length of the data read as well
// as the source address. We're not interested in the length, but we need to keep
// track of the source in order to send our reply later on.
// Next, `DnsPacket::from_buffer` is used to parse the raw bytes into
// a `DnsPacket`. It uses the same error handling idiom as the previous statement.
let request = match DnsPacket::from_buffer(&mut req_buffer) {
Ok(x) => x,
Err(e) => {
println!("Failed to parse UDP query packet: {:?}", e);
continue;
}
};
// Create and initialize the response packet
let mut packet = DnsPacket::new();
packet.header.id = request.header.id;
packet.header.recursion_desired = true;
packet.header.recursion_available = true;
packet.header.response = true;
// Being mindful of how unreliable input data from arbitrary senders can be, we
// need make sure that a question is actually present. If not, we return `FORMERR`
// to indicate that the sender made something wrong.
if request.questions.is_empty() {
packet.header.rescode = ResultCode::FORMERR;
}
// Usually a question will be present, though.
else {
let question = &request.questions[0];
println!("Received query: {:?}", question);
// Since all is set up and as expected, the query can be forwarded to the target
// server. There's always the possibility that the query will fail, in which case
// the `SERVFAIL` response code is set to indicate as much to the client. If
// rather everything goes as planned, the question and response records as copied
// into our response packet.
if let Ok(result) = lookup(&question.name, question.qtype, server) {
packet.questions.push(question.clone());
packet.header.rescode = result.header.rescode;
for rec in result.answers {
println!("Answer: {:?}", rec);
packet.answers.push(rec);
}
for rec in result.authorities {
println!("Authority: {:?}", rec);
packet.authorities.push(rec);
}
for rec in result.resources {
println!("Resource: {:?}", rec);
packet.resources.push(rec);
}
} else {
packet.header.rescode = ResultCode::SERVFAIL;
}
// The only thing remaining is to encode our response and send it off!
let mut res_buffer = BytePacketBuffer::new();
match packet.write(&mut res_buffer) {
Ok(_) => {},
Err(e) => {
println!("Failed to encode UDP response packet: {:?}", e);
continue;
}
};
let len = res_buffer.pos();
let data = match res_buffer.get_range(0, len) {
Ok(x) => x,
Err(e) => {
println!("Failed to retrieve response buffer: {:?}", e);
continue;
}
};
match socket.send_to(data, src) {
Ok(_) => {},
Err(e) => {
println!("Failed to send response buffer: {:?}", e);
continue;
}
};
}
} // End of request loop
} // End of main
The match idiom for error handling is used again and again here, since we want to avoid
terminating our request loop at all cost. It's a bit verbose, and normally we'd
like to use try! instead. Unfortunately that's unavailable to us here, since
we're in the main function which doesn't return a Result.
All done! Let's try it! We start our server in one terminal, and use dig to
perform a lookup in a second terminal.
# dig @127.0.0.1 -p 2053 google.com
; <<>> DiG 9.10.3-P4-Ubuntu <<>> @127.0.0.1 -p 2053 google.com
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 47200
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0
;; QUESTION SECTION:
;google.com. IN A
;; ANSWER SECTION:
google.com. 68 IN A 216.58.211.142
;; Query time: 1 msec
;; SERVER: 127.0.0.1#2053(127.0.0.1)
;; WHEN: Fri Jul 08 12:07:44 CEST 2016
;; MSG SIZE rcvd: 54
Looking at our server terminal we see:
Received query: DnsQuestion { name: "google.com", qtype: A }
Answer: A { domain: "google.com", addr: 216.58.211.142, ttl: 96 }
Success! In less than 800 lines of code, we've built a DNS server able to respond to queries with several different record types!
In the next chapter, we'll get rid of our dependence on an existing resolver: Chapter 5 - Recursive Resolve