Crate proc_sys_parser

This crate provides routines for parsing linux /proc files into Rust structs.

There are multiple other crates doing this, but these either do not choose to process the statistics in way to make them directly usable, or generalize the statistics and loose the detail.

Currently, only two /proc files are processed:

/proc/stat

The processor of /proc/stat reads the CLK_TCK setting and transforms the jiffies of the cpu times into milliseconds.

Here is an example using /proc/stat:

use proc_sys_parser::{stat, stat::{ProcStat, CpuStat}};

let proc_stat = stat::read();

assert_eq!(proc_stat,
    ProcStat { cpu_total: CpuStat { name: "cpu".to_string(), user: 1807810, nice: 3570, system: 1993480, idle: 1571076180, iowait: 106510, irq: 0, softirq: 22630, steal: 0, guest: 0, guest_nice: 0, },
               cpu_individual: vec![ CpuStat { name: "cpu0".to_string(), user: 291210, nice: 0, system: 346690, idle: 261693790, iowait: 17710, irq: 0, softirq: 17260, steal: 0, guest: 0, guest_nice: 0, },
                                     CpuStat { name: "cpu1".to_string(), user: 296120, nice: 640, system: 333930, idle: 261871750, iowait: 18610, irq: 0, softirq: 1130, steal: 0, guest: 0, guest_nice: 0, },
                                     CpuStat { name: "cpu2".to_string(), user: 293240, nice: 470, system: 329740, idle: 261885790, iowait: 16920, irq: 0, softirq: 1140, steal: 0, guest: 0, guest_nice: 0, },
                                     CpuStat { name: "cpu3".to_string(), user: 312120, nice: 2450, system: 329940, idle: 261867390, iowait: 18590, irq: 0, softirq: 1210, steal: 0, guest: 0, guest_nice: 0, },
                                     CpuStat { name: "cpu4".to_string(), user: 306320, nice: 0, system: 327210, idle: 261879650, iowait: 17120, irq: 0, softirq: 820, steal: 0, guest: 0, guest_nice: 0, },
                                     CpuStat { name: "cpu5".to_string(), user: 308780, nice: 0, system: 325950, idle: 261877780, iowait: 17540, irq: 0, softirq: 1040, steal: 0, guest: 0, guest_nice: 0, },
                                    ],
               interrupts: vec![ 64226374, 0, 1578053, 20277276, 0, 0, 0, 2, 0, 0, 0, 38423958, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 644, 0, 0, 0, 0, 0, 2, 0, 82584, 85616, 82930, 74304, 74994, 81970, 0, 0, 0, 22407, 22081, 21799, 21485, 22304, 21123, 0, 1186950, 860525, 0, 0, 4295, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1281072, 0, 0, ],
               context_switches: 107129963,
               boot_time: 1701783048,
               processes: 1118791,
               processes_running: 1,
               processes_blocked: 0,
               softirq: vec![ 22446125, 32, 4737167, 213, 2012904, 11, 0, 1450, 7408871, 0, 8285477, ],
             });

/proc/schedstat

The processor of /proc/schedstat also reads the CLK_TCK setting and transforms the jiffies with the cpu fields, which are fields 7 (time spent running by tasks on this processor) and 8 (time spent waiting to run by tasks on this processor) to milliseconds.

Here is an example using /proc/schedstat:

use proc_sys_parser::{schedstat, schedstat::ProcSchedStat};

let proc_schedstat = schedstat::read();

assert_eq!(proc_schedstat,
ProcSchedStat { version: 15,
                timestamp: 4318961659,
                cpu: vec![
                   vec![0, 0, 0, 0, 0, 0, 0, 4575719016330, 485940746140, 4348645],
                   vec![1, 0, 0, 0, 0, 0, 0, 4352064330120, 449441457150, 3928368],
                   vec![2, 0, 0, 0, 0, 0, 0, 4296375140810, 435916732570, 3833297],
                   vec![3, 0, 0, 0, 0, 0, 0, 4453083890360, 431027439820, 3851418],
                   vec![4, 0, 0, 0, 0, 0, 0, 4386665545210, 437068452780, 3787400],
                   vec![5, 0, 0, 0, 0, 0, 0, 4447083238720, 428623717880, 3900565]],
                domain: vec![
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
                   vec![0, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]
        });

cpu vector

!! Please mind that the vector with cpu statistics takes the cpu number as the first field in the vector.

This means that the fields 7 (running on cpu time) and 8 (waiting on cpu runtime) are fields 8 and 9 in the vector.

domain vector

!!Please mind that the vector with domain statistics takes the domain number as the first field in the vector, and the cpumask as the second field.

This means the numbers with the description for the fields in the kernel documentation https://www.kernel.org/doc/Documentation/scheduler/sched-stats.txt has to be increased by two to get the right statistic number in the vector.

Modules

• schedstat
• stat