Understanding Linux poll(): Structure, Events, and How It Works

What is poll()

poll() is a Linux system call used for I/O multiplexing, allowing a program to monitor multiple file descriptors for readiness events.

Function prototype

<code>struct pollfd myfd[500];
int poll(struct pollfd *fds, nfds_t nfds, int timeout);
</code>

pollfd structure

<code>struct pollfd {
    int fd;          // file descriptor to monitor
    short events;    // requested events
    short revents;   // returned events
};
</code>

Event flags

pollfd.events and pollfd.revents can be set to the following flags:

Read events: POLLIN, POLLRDNORM, POLLRDBAND, POLLRI

Write events: POLLOUT, POLLWRNORM, POLLWRBAND

Error events: POLLERR, POLLHUP, POLLNVAL

Multiple flags can be combined using the bitwise OR operator, e.g., POLLIN | POLLOUT .

Parameters

nfds : the highest-numbered file descriptor in the pollfd array plus one.

timeout : timeout in milliseconds; >0 waits up to the specified time, =0 returns immediately, =-1 blocks indefinitely.

Return value

>0: number of file descriptors ready.

-1: error occurred.

How poll works

The typical workflow is:

Application fills a pollfd array with file descriptors and the events it wants to monitor.

Calling poll() copies this array into the kernel.

The kernel monitors the descriptors; when data arrives, DMA copies it to the kernel buffer and updates the corresponding revents field.

poll() returns; the application checks the return value and iterates the pollfd array to handle ready descriptors.

Key differences from select

poll introduces the pollfd structure, removing the 1024‑descriptor limit of select on 32‑bit systems and allowing the structure to be reused. However, poll is Linux‑specific, while select works on many platforms.