Download PDF

Chapter 35G: Networking, time, and logging

The rootfs boots. Your app starts. Now the normal field problems begin:

  • It has no IP address.

  • DNS does not work.

  • Time is 1970.

  • Logs vanish after reboot.

  • The network comes up after your app already failed.

This chapter gives the practical checklist.

35G.1 Bring up Ethernet manually

On a small BusyBox rootfs:

[root@pa-mini:~]# ip link set eth0 up
[root@pa-mini:~]# ip addr add 192.168.1.50/24 dev eth0
[root@pa-mini:~]# ip route add default via 192.168.1.1
[root@pa-mini:~]# echo 'nameserver 8.8.8.8' > /etc/resolv.conf

Test:

[root@pa-mini:~]# ping -c 3 192.168.1.1
[root@pa-mini:~]# ping -c 3 8.8.8.8
[root@pa-mini:~]# ping -c 3 example.com

If the first works but the second fails, routing is wrong.

If the second works but the third fails, DNS is wrong.

35G.2 DHCP with BusyBox

BusyBox provides udhcpc.

[root@pa-mini:~]# udhcpc -i eth0

For boot, add to /etc/init.d/rcS:

ip link set eth0 up
udhcpc -i eth0 -q

-q exits after it gets a lease. Without it, udhcpc stays in the foreground unless scripted differently.

35G.3 Static IP config file

For a simple product, keep one config file:

# /etc/network.conf
IFACE=eth0
ADDR=192.168.1.50/24
GATEWAY=192.168.1.1
DNS=192.168.1.1

Read it from rcS:

if [ -f /etc/network.conf ]; then
    . /etc/network.conf
    ip link set "$IFACE" up
    ip addr add "$ADDR" dev "$IFACE"
    ip route add default via "$GATEWAY"
    echo "nameserver $DNS" > /etc/resolv.conf
else
    udhcpc -i eth0 -q
fi

Move the runtime-editable copy to /data/network.conf if rootfs is read-only.

35G.4 Hostname and /etc/hosts

Set:

echo pa-mini > /etc/hostname

And:

127.0.0.1   localhost
127.0.1.1   pa-mini

On boot:

hostname -F /etc/hostname

Some tools behave strangely when hostname is missing.

35G.5 Time

Many i.MX6ULL boards do not have a battery-backed RTC installed. Without network time, the system may boot at 1970.

Check:

[root@pa-mini:~]# date

Use BusyBox NTP:

[root@pa-mini:~]# ntpd -n -q -p pool.ntp.org

For boot:

ntpd -p pool.ntp.org

On Ubuntu-base, use systemd-timesyncd or chrony:

[root@pa-mini:~]# timedatectl status
[root@pa-mini:~]# systemctl status systemd-timesyncd

If the device has no internet, set time from:

  • RTC,

  • GPS,

  • local NTP server,

  • manufacturing tool,

  • signed update metadata.

35G.6 Logging paths

Decide where logs live:

Path

Storage

Use

/run/log

tmpfs

boot-only logs

/var/log

tmpfs or persistent

system logs

/data/log

persistent data partition

field logs

journalctl

systemd journal

Ubuntu/systemd images

For BusyBox syslog:

syslogd -n -O /var/log/messages

For background:

syslogd -O /var/log/messages
klogd

If /var/log is tmpfs, logs are lost on reboot. That is okay for development. For field debugging, copy important logs to /data/log.

35G.7 Log rotation

Without rotation, logs fill storage.

Simple BusyBox rotation:

LOG=/data/log/messages

if [ -f "$LOG" ] && [ "$(wc -c < "$LOG")" -gt 1048576 ]; then
    mv "$LOG" "$LOG.1"
    : > "$LOG"
fi

Run it from cron or before starting syslog.

Buildroot can also include logrotate.

35G.8 Debug checklist

When the board says “network broken”:

ip link
ip addr
ip route
cat /etc/resolv.conf
ping -c 3 <gateway>
ping -c 3 8.8.8.8
ping -c 3 example.com
dmesg | grep -i eth

When time is wrong:

date
hwclock -r
timedatectl status 2>/dev/null || true
ps | grep ntp

When logs are missing:

mount | grep ' /var\| /data'
ps | grep 'syslog\|journald'
ls -lh /var/log /data/log 2>/dev/null

35G.9 Key Notes

  • Link state, IP configuration, routing and DNS are separate layers. Success at one layer narrows a diagnosis but does not establish that the next layer works.

  • Network configuration needs one clear owner. Competing static settings, DHCP clients and service scripts can overwrite addresses or resolver state.

  • Wall-clock time supports certificates and useful logs, while elapsed-time measurements need a clock that is not changed by synchronization.

  • Log retention is a storage policy. Tmpfs limits persistence, while flash-backed logs need capacity and wear controls.

  • A service should cope with delayed or lost connectivity. One successful boot-time connection is not a lifetime availability guarantee.

35G.10 Pitfalls

  • App assumes network is ready. Network may come up after init starts your app. Retry inside the app.

  • DNS file overwritten. DHCP clients may replace /etc/resolv.conf.

  • No RTC. TLS, package managers, and certificates fail if time is far wrong.

  • Persistent logs fill flash. Rotate or cap them.

  • Logging too much to eMMC. Flash has wear. Use tmpfs for noisy logs and persistent storage only for useful summaries.

Previous chapter: Chapter 35F - Your application as a service

Next chapter: Chapter 35B - Read-only rootfs and overlayfs