How Does htop Read Metrics From /proc?

The htop command-line utility is a staple for Linux system administrators, offering a real-time, interactive overview of system resources like CPU usage, memory consumption, and running processes. Unlike static tools, htop achieves its dynamic monitoring by continuously parsing virtual files within the /proc filesystem—a pseudo-filesystem generated on the fly by the Linux kernel. This article explores how htop translates these raw, text-based kernel metrics into an intuitive, color-coded terminal interface.

The Pseudo-Filesystem as a Data Source

In Linux, the /proc directory does not contain actual files stored on a disk; instead, it acts as an interface to the Linux kernel’s internal data structures. When htop requests information, the kernel generates text outputs representing the current state of the system.

htop specifically targets several key files to populate its main dashboard:

/proc/stat: This file provides aggregate kernel statistics. htop reads this to calculate overall and per-core CPU utilization. It extracts time spent in various states, such as user, system, nice, and idle.
/proc/meminfo: This contains detailed statistics about system memory. htop parses this file to display total, used, free, and buffered/cached RAM, as well as swap usage.
/proc/loadavg: This file provides the system load averages over 1, 5, and 15-minute intervals, which htop displays in the upper-right corner of its interface.

Tracking Individual Processes

To populate the detailed process list at the bottom of the screen, htop scans the numbered directories within /proc. Each number corresponds to a specific Process ID (PID) currently running on the system.

For every PID directory, htop opens and reads several specific files:

/proc/[PID]/stat: Contains status information about the process, such as its current state (running, sleeping, zombie), virtual memory size, and CPU time consumed in user and kernel mode.
/proc/[PID]/statm: Provides a simplified breakdown of the process’s memory usage in pages, allowing htop to determine resident set size (RSS) and shared memory.
/proc/[PID]/cmdline: Holds the command that originally started the process, which htop formats and displays in the “Command” column.
/proc/[PID]/status: Contains a human-readable version of much of the data in stat, including user and group IDs (UID/GID) used to display the process owner.

The Read-and-Refresh Cycle

Because system metrics change every millisecond, htop operates in a continuous loop dictated by its refresh interval (defaulting to every 1.5 seconds). During each cycle, the utility clears its internal data structures, performs rapid read operations across the necessary /proc files, and calculates differences between the current data and the data from the previous cycle.

For example, to calculate the CPU percentage shown in the bars, htop cannot rely on a single snapshot from /proc/stat. It must take two reads separated by time, calculate the delta of total CPU time versus idle CPU time, and convert that ratio into a percentage. Once these calculations are complete, htop uses the ncurses library to efficiently redraw only the parts of the terminal screen that have changed, ensuring minimal CPU overhead from the monitoring tool itself.