cabletest/counters.go

package main

import (
	"os"
	"path/filepath"
	"strconv"
	"strings"
)

// Split by direction of travel so each counter belongs to exactly one
// direction: the transmitting interface owns the tx fields and the receiving one
// the rx fields. Reading both sets off both interfaces double counts every drop.
//
// The two sides come from different places, and deliberately not from both at
// once. dev_get_stats adds the core's own drop counts on top of whatever the
// driver reports, so a sysfs field is the driver's counter plus the stack's and
// overlaps any driver counter read beside it.
//
// Sending is the stack's story: what it refused, and what the qdisc discarded
// because we outran it, neither of which the driver ever sees.
var nicTxFields = []string{
	"tx_errors", "tx_dropped", "tx_fifo_errors",
	"tx_carrier_errors", "tx_aborted_errors", "tx_window_errors",
	"collisions",
}

// Receiving is the hardware's, and is taken from the driver's own array:
// sysfs flattens and overlaps it (crc errors fold into rx_errors, so summing
// rx_errors alongside rx_crc_errors charges every bad frame twice). Named
// individually, so nothing contains anything else in the list; the 82599
// exposes no jabber, fragment, illegal-byte or fault counters (docs/nics/x520/).
var nicRxFields = []string{
	"rx_crc_errors",
	"rx_missed_errors",
	"rx_length_errors",
	"rx_long_length_errors",
	"rx_short_length_errors",
}

func ifDriver(name string) (string, error) {
	link, err := os.Readlink("/sys/class/net/" + name + "/device/driver")
	if err != nil {
		return "", err
	}
	return filepath.Base(link), nil
}

func readUint(path string) (uint64, bool) {
	b, err := os.ReadFile(path)
	if err != nil {
		return 0, false
	}
	v, err := strconv.ParseUint(strings.TrimSpace(string(b)), 10, 64)
	if err != nil {
		return 0, false
	}
	return v, true
}

// Every counter here was read once before the test began, so one that stops
// reading now is the interface going away underneath us. Stopping is the only
// honest answer: a tester that quietly gives up counting errors goes on
// reporting a clean link, which is the one lie it must never tell.
func mustReadUint(path string) uint64 {
	v, ok := readUint(path)
	if !ok {
		panic("reading " + path)
	}
	return v
}

func sumFields(base string, fields []string) uint64 {
	var total uint64
	for _, f := range fields {
		total += mustReadUint(base + "/statistics/" + f)
	}
	return total
}

// A read is an ioctl and a handful of sysfs attributes, each costing
// microseconds of open and read, so it rides the sampler's goroutine and
// clock rather than the one that has a frame to draw every sixteen
// milliseconds; a link error lands on the panel as quickly as a lost frame.
type nicCounters struct {
	txBase string
	rx     *statReader
	raw    uint64
}

func newNICCounters(fd int, txName, rxName string) (*nicCounters, error) {
	rx, err := newStatReader(fd, rxName, nicRxFields)
	if err != nil {
		return nil, err
	}
	return &nicCounters{
		txBase: "/sys/class/net/" + txName,
		rx:     rx,
	}, nil
}

// Carrier is a property of the netdev rather than a counter of either side, and
// is taken from the transmitting interface only: across both directions each
// interface transmits exactly once, so every flap is counted once.
func (c *nicCounters) read() uint64 {
	return sumFields(c.txBase, nicTxFields) + c.rx.sum() +
		mustReadUint(c.txBase+"/carrier_down_count")
}

// Nic counters run from boot and restart from zero whenever the driver resets
// its statistics, so only forward motion is the delta: a raw difference would
// charge a boot's worth of errors to the first collection and turn a reset
// into a near-2^64 underflow.
func (c *nicCounters) delta() uint64 {
	raw := c.read()
	var d uint64
	if raw > c.raw {
		d = raw - c.raw
	}
	c.raw = raw
	return d
}