Every service that communicates over the network uses a port number. Instead of memorising port numbers, Linux stores a mapping of service names to port numbers in a plain text file: /etc/services.

Format of /etc/services

Each line in this file looks like this:

# service-name   port/protocol   [aliases...]   [# comment]

ssh             22/tcp
ssh             22/udp
http            80/tcp    www     # World Wide Web HTTP
https           443/tcp
ntp             123/udp           # Network Time Protocol
syslog          514/udp
rsh             514/tcp           # Remote Shell

Each line has:

• service-name – human-readable name (e.g., “ssh”, “http”)
• port/protocol – the port number followed by the protocol (tcp or udp)
• aliases – optional alternative names for the service
• # comment – optional explanation

Visualising the /etc/services Lookup

Important Rules About Port Assignments

IANA (Internet Assigned Numbers Authority) is the organisation that officially assigns port numbers. Here are some important rules to understand:

Linux provides library functions to look up the /etc/services file programmatically. The classic functions are getservbyname() and getservbyport(). These are older APIs (single-threaded, not reentrant), but useful to understand.

getservbyname() — Look up a service by name

#include <netdb.h>
#include <stdio.h>
#include <arpa/inet.h>

int main() {
    struct servent *se;

    /* Look up the "http" service for TCP */
    se = getservbyname("http", "tcp");
    if (se == NULL) {
        fprintf(stderr, "Service not found\n");
        return 1;
    }

    printf("Service name : %s\n", se->s_name);
    printf("Port number  : %d\n", ntohs(se->s_port));  /* ntohs: network to host byte order */
    printf("Protocol     : %s\n", se->s_proto);

    return 0;
}

Output:

Service name : http
Port number  : 80
Protocol     : tcp

getservbyport() — Look up a service by port number

#include <netdb.h>
#include <stdio.h>
#include <arpa/inet.h>

int main() {
    struct servent *se;

    /* Port 22 in network byte order */
    se = getservbyport(htons(22), "tcp");
    if (se == NULL) {
        fprintf(stderr, "Port not found in /etc/services\n");
        return 1;
    }

    printf("Service name : %s\n", se->s_name);
    printf("Port number  : %d\n", ntohs(se->s_port));
    printf("Protocol     : %s\n", se->s_proto);

    return 0;
}

Output:

Service name : ssh
Port number  : 22
Protocol     : tcp

Old socket programs were written specifically for IPv4. The functions they used (like gethostbyname()) returned only IPv4 addresses. As IPv6 grew, programmers had to rewrite their code to handle IPv6 separately.

The modern solution is to use getaddrinfo() and getnameinfo() — they handle both IPv4 and IPv6 transparently. Your code doesn’t need to know which protocol is being used.

Old Way vs New Way

Flow Diagram: Old vs New

Functions like gethostbyname() use a static internal buffer to store their result. This means if two threads call the function at the same time, they will corrupt each other’s results. Such functions are called non-reentrant.

Q1. What is the /etc/services file and what is its purpose?

Answer: /etc/services is a plain text configuration file in Linux that provides a mapping between human-readable service names (like “http”, “ssh”) and their corresponding port numbers and protocols (tcp or udp). It helps programmers and system administrators refer to services by name rather than memorising port numbers. The C library functions getservbyname() and getservbyport() read this file to perform lookups.

Q2. Is /etc/services a port reservation mechanism?

Answer: No. /etc/services is purely a name-to-number mapping reference. Appearing in this file does not guarantee that the port is available. Another process could already be bound to that port. If you try to bind to a port that is already in use, bind() will fail with EADDRINUSE.

Q3. Why does IANA assign both TCP and UDP ports even if only one protocol is used?

Answer: This is IANA policy for consistency and future-proofing. Even if a service currently uses only TCP (like HTTP), the UDP port is reserved under the same number to prevent conflicts. This avoids a situation where someone assigns UDP 80 to a completely different service in the future, which would cause confusion.

Q4. Why is gethostbyname() considered obsolete?

Answer: Two main reasons: (1) It only supports IPv4, not IPv6. (2) It uses a static internal buffer, making it non-reentrant and unsafe in multi-threaded programs. The modern replacement is getaddrinfo(), which is reentrant, supports both IPv4 and IPv6, and combines hostname and service resolution in a single call.

Q5. What is the difference between TCP port 514 and UDP port 514?

Answer: They are assigned to different services — TCP 514 is used by rsh (remote shell), while UDP 514 is used by the syslog daemon. This is an exception to the normal IANA policy (where both protocols get the same service) and exists because these assignments were made before the current IANA policy was adopted.

Q6. Name two services that use both TCP and UDP on the same port.

Answer: DNS (port 53) and echo (port 7) are examples of services that actively use both TCP and UDP.

Q7. What does “protocol-independent programming” mean in the context of sockets?

Answer: It means writing network code that works transparently with both IPv4 and IPv6 without needing separate code paths. Using getaddrinfo() and getnameinfo() achieves this because they return/accept generic socket address structures that can represent either protocol, letting the OS handle the protocol-specific details.