NetIO

NetIO

Print network I/O throughput

Module type	netio
Default order	67 (only used by --gen-config)
Module source	src/modules/netio/netio.c
Detection source	src/detection/netio/

Prints one line per network interface with the data received and sent since the previous snapshot, divided by the elapsed time — a rate in bytes per second. By default only the interface that carries the default route is shown. The key is Network I/O (<interface name>).

Network I/O (en0): 944 B/s (IN) - 849 B/s (OUT)
Network I/O (lo0): 361 B/s (IN) - 361 B/s (OUT)

(IN) is received data and (OUT) is sent data. When defaultRouteOnly is turned off, the interface that carries the default route is marked with a trailing *. With detectTotal: true the module prints the cumulative counters since boot and the /s suffix disappears.

Like Modules/DiskIO, this module needs two readings to compute a rate and therefore waits waitTime milliseconds (250 by default) before it can print anything. detectTotal: true does not wait.

Platform support

Platform	Implementation	Notes
Linux	netio_linux.c	/sys/class/net/*/statistics, honouring operstate
Android	netio_linux.c	Same file
GNU/Hurd	netio_nosupport.c	Reports "Not supported on this platform"
FreeBSD	netio_bsd.c	sysctl(NET_RT_IFLIST)
NetBSD	netio_bsd.c	Same file
OpenBSD	netio_bsd.c	Same file
Solaris / illumos	netio_sunos.c	kstat chain, class net, module link
Haiku	netio_haiku.cpp	BNetworkRoster plus GetStats()
macOS	netio_apple.c	sysctl(IFMIB_IFDATA)
Windows	netio_windows.c	GetAdaptersAddresses() plus GetIfEntry2()

GNU/Hurd is the only platform without an implementation.

Configuration

Key	Type	Default	Description
namePrefix	string	empty	Only show interfaces whose name starts with this string
defaultRouteOnly	boolean	true (false on Android)	Show only the interface used for the default route
detectTotal	boolean	false	Print cumulative counters instead of a per-second rate
waitTime	integer	250	Milliseconds between the two counter readings. Ignored when detectTotal is true.
key	string	Network I/O (<ifname>)	Module key. A single space hides the key and the separator.
keyColor	color	–	Overrides display.color.keys
keyIcon	string	built-in glyph	Printed when display.key.type includes the icon bit. Any glyph works; "" prints none.
keyWidth	integer	–	Overrides display.key.width
outputColor	color	–	Overrides display.color.output
format	string	–	Custom output format (see below)
condition	object	–	Show the module only if the conditions match

waitTime accepts 0…4294967295, but 0 is silently raised to 1; anything outside the range — including a negative number — is rejected with a message instead of being reinterpreted.

The Android default for defaultRouteOnly is false because the interface that the option filters against is not available through the same API there; on every other platform the option behaves as documented.

Format string

Run fastfetch -h netio-format for the authoritative list.

Variable	Description
{ifname}	Interface name — also available in key format
{rx-size}	Received bytes, formatted with display.size, plus /s unless detectTotal
{tx-size}	Sent bytes, formatted the same way
{rx-bytes}	Bytes received per second, as a raw number
{tx-bytes}	Bytes sent per second, as a raw number
{rx-packets}	Packets received per second
{tx-packets}	Packets sent per second
{rx-errors}	Receive errors per second
{tx-errors}	Send errors per second
{rx-drops}	Packets dropped while receiving, per second
{tx-drops}	Packets dropped while sending, per second
{is-default-route}	true when the interface carries the default route

The …-bytes, …-packets, …-errors and …-drops values are already rates; only {rx-size} and {tx-size} carry a /s suffix of their own. {is-default-route} is about the interface, not about the current configuration — it is true for the default route even when defaultRouteOnly is on and the * marker is suppressed.

JSON output

{
    "type": "NetIO",
    "result": [
        {
            "name": "en0",
            "defaultRoute": true,
            "txBytes": 0,
            "rxBytes": 262,
            "txPackets": 0,
            "rxPackets": 3,
            "rxErrors": 0,
            "txErrors": 0,
            "rxDrops": 0,
            "txDrops": 0
        }
    ]
}

The ten numeric fields are per-second rates unless detectTotal is set, in which case they are the raw cumulative counters. On failure the object is { "type": "NetIO", "error": "…" }.

Examples

// Every interface, with the default route marked
{ "type": "netio", "defaultRouteOnly": false }

// Cumulative counters, no waiting
{ "type": "netio", "detectTotal": true }

// A Wi-Fi-only line, refreshed over one second
{ "type": "netio", "namePrefix": "wlan", "defaultRouteOnly": false, "waitTime": 1000 }

// Compact upload/download line
{ "type": "netio", "format": "IN {rx-size} OUT {tx-size}" }

Pitfalls

• The module blocks for waitTime milliseconds. The default 250 ms is paid by every run that uses the rate path, including --format json. A larger value really does block — 4294967295 is about 49.7 days and is accepted without a warning — so a mistyped waitTime looks like a hang.
• Only one interface is shown by default. defaultRouteOnly is true, which is why a laptop prints a single line even though a dozen interfaces exist. Turn it off to see the rest; the default route is then marked with *.
• The * marker and {is-default-route} do not agree in the default configuration. The marker is suppressed whenever defaultRouteOnly is on, while the variable keeps reporting true; a format that renders both shows the variable but no star.
• Linux hides interfaces that are not up. Anything whose operstate is not up or unknown is skipped, so a Wi-Fi interface that is associated but administratively down, or a USB Ethernet adapter with no link, is missing from the list rather than reported as zero. macOS, FreeBSD, NetBSD and OpenBSD filter on IFF_RUNNING and skip IFF_NOARP interfaces instead.
• Windows does not filter at all. Every adapter returned by GetAdaptersAddresses() is listed — disconnected Wi-Fi, VPN tunnels, Hyper-V switches — when defaultRouteOnly is off, so the interface count depends much more on the platform than on the machine.
• Two snapshots can disagree. An interface appearing or disappearing between them aborts the module with Different number of network interfaces. Network change?; a renamed interface aborts it with Network interface name changed. Both messages only appear when display.showErrors is true.
• namePrefix is matched against the platform's own interface naming. On Linux and Solaris that is the kernel name (eth0, en0, net0), on macOS the BSD name, and on Windows the adapters' friendly name (Ethernet, Wi-Fi) — so the same value does not port between platforms. A prefix that matches nothing leaves the module with an empty list, and it then reports No network interfaces found.
• {tx-drops} is always zero on Solaris and on older BSDs. The Solaris kstat interface has no drop counter, and BSD builds only read ifi_oqdrops when the platform provides it; the field is filled with 0 rather than left out of the JSON result.
• The JSON array is not sorted, the text output is. The sort by name happens while printing, so --format json returns the detection order — readdir() order for Linux, the sysctl order for macOS and the BSDs.
• Under --dynamic-interval the default route is re-resolved every round. The netif layer that supplies it is invalidated at each round boundary, so moving between Wi-Fi and Ethernet makes the reported interface follow — and because the module then sees a different interface set, a round can fail with "Different number of network interfaces" instead of printing a rate.
• Every option warning is printed twice when the module is configured from a file. The module object is parsed once in the prepare pass and once while printing, and the parser reports during both: a config with one unknown key and one out-of-range waitTime produces four lines, not two. The same doubling happens with -s netio as long as a config supplies the options.

Implementation

Every backend implements ffNetIOGetIoCounters(). Rates are computed in the shared src/detection/netio/netio.c: ffPrepareNetIO() stores the first snapshot in a file-scope static, ffDetectNetIO() takes the second after sleeping until waitTime has elapsed and rewrites each counter as (current − previous) × 1000 / elapsed_ms, keeping the raw value as the next baseline. The eight uint64_t counters are walked in one loop, guarded by a static_assert on the struct layout. The snapshots are matched positionally, which is why a changed interface count or a renamed interface is an error rather than a skipped entry.

All ten backends call ffNetifGetDefaultRouteV4() first: with defaultRouteOnly they fetch only that interface, and without it they use the result to set the flag behind * and {is-default-route}.

Linux and Android

/sys/class/net/ is walked and each interface is read through directory descriptors: operstate gates the entry, then statistics/rx_bytes, tx_bytes, rx_packets, tx_packets, rx_errors, tx_errors, rx_dropped and tx_dropped are parsed. Only up and unknown pass the gate — unknown is how the loopback interface reports itself, which is why lo is listed — while down, lowerlayerdown and dormant interfaces are dropped.

FreeBSD, NetBSD, OpenBSD and macOS

sysctl(NET_RT_IFLIST) with the default route's interface index when defaultRouteOnly is set, and with 0 otherwise. The if_msghdr chain is walked, skipping anything that is not RTM_IFINFO, not IFF_RUNNING or flagged IFF_NOARP; the interface name comes from the following sockaddr_dl. The counters are ifi_ibytes/ifi_obytes, ifi_ipackets/ifi_opackets, ifi_ierrors/ifi_oerrors and the drop counters, with txDrops only available when the build has ifi_oqdrops.

macOS uses the same mib through the if_mib layer (IFMIB_IFDATA for one interface, IFMIB_IFALLDATA for all of them) and reads ifmibdata records, including ifmd_snd_drops for the send-side drops.

Solaris and illumos

The kstat chain is walked for entries whose class is net and whose module is link; the name is the kstat name and the default route is compared by name. obytes64, rbytes64, opackets64, ipackets64, oerrors and ierrors are looked up with kstat_data_lookup(), and both drop counters are set to 0 because the interface has no such statistic.

Haiku

BNetworkRoster::Default() enumerates the interfaces, interface.Exists() filters the unused ones and GetStats() fills an ifreq_stats with send/receive byte, packet, error and drop counters. The default route is identified by interface index.

Windows

GetAdaptersAddresses() is called in a retry loop (up to four times, growing the buffer on ERROR_BUFFER_OVERFLOW) with anycast, multicast and DNS lookups skipped. The friendly name is converted from UTF-16 and GetIfEntry2() supplies the MIB_IF_ROW2 counters: InOctets/OutOctets, unicast plus non-unicast packets, InErrors/OutErrors and InDiscards/OutDiscards. Unlike the BSD path there is no link-state filter, so adapters with no link are listed with counters of their own.