systemd: pivot WantedBy to interface rather than multi-user

[jimdigriz]

Fixes SQM restarts (issue #143) when the interface is removed and re-added (eg. PPP) and also SQM loads sooner rather than waiting for multi-user to start

[tohojo]

So did you get this working for manual SIGTERM of pppd as well? :)

[jimdigriz]

Yes, Works for Me(tm) :)

root@clearfog:/home/alex# systemctl status --no-pager -n0 [email protected]
● [email protected] - SQM scripts for iface wan
     Loaded: loaded (/lib/systemd/system/[email protected]; enabled; vendor preset: enabled)
     Active: active (exited) since Tue 2022-01-11 17:11:15 UTC; 31min ago
    Process: 4452 ExecStart=/usr/lib/sqm/start-sqm (code=exited, status=0/SUCCESS)
   Main PID: 4452 (code=exited, status=0/SUCCESS)
        CPU: 164ms

root@clearfog:/home/alex# systemctl status --no-pager -n0 [email protected]
● [email protected] - PPP connection for wan
     Loaded: loaded (/etc/systemd/system/[email protected]; enabled; vendor preset: enabled)
     Active: active (running) since Tue 2022-01-11 17:11:18 UTC; 31min ago
       Docs: man:pppd(8)
    Process: 4428 ExecStartPre=/lib/systemd/systemd-networkd-wait-online -i eth1 -o carrier (code=exited, status=0/SUCCESS)
    Process: 4439 ExecStart=/usr/sbin/pppd plugin rp-pppoe.so eth1 call wan linkname eth1 ifname wan updetach (code=exited, status=0/SUCCESS)
   Main PID: 4588 (pppd)
      Tasks: 1 (limit: 4169)
     Memory: 884.0K
        CPU: 107ms
     CGroup: /system.slice/system-pppd\x2deth1.slice/[email protected]
             └─4588 /usr/sbin/pppd plugin rp-pppoe.so eth1 call wan linkname eth1 ifname wan updetach

root@clearfog:/home/alex# kill 4588

root@clearfog:/home/alex# systemctl status --no-pager -n0 [email protected]
● [email protected] - PPP connection for wan
     Loaded: loaded (/etc/systemd/system/[email protected]; enabled; vendor preset: enabled)
     Active: active (running) since Tue 2022-01-11 17:43:11 UTC; 1s ago
       Docs: man:pppd(8)
    Process: 4934 ExecStartPre=/lib/systemd/systemd-networkd-wait-online -i eth1 -o carrier (code=exited, status=0/SUCCESS)
    Process: 4935 ExecStart=/usr/sbin/pppd plugin rp-pppoe.so eth1 call wan linkname eth1 ifname wan updetach (code=exited, status=0/SUCCESS)
   Main PID: 5081 (pppd)
      Tasks: 1 (limit: 4169)
     Memory: 892.0K
        CPU: 72ms
     CGroup: /system.slice/system-pppd\x2deth1.slice/[email protected]
             └─5081 /usr/sbin/pppd plugin rp-pppoe.so eth1 call wan linkname eth1 ifname wan updetach

root@clearfog:/home/alex# systemctl status --no-pager -n0 [email protected]
● [email protected] - SQM scripts for iface wan
     Loaded: loaded (/lib/systemd/system/[email protected]; enabled; vendor preset: enabled)
     Active: active (exited) since Tue 2022-01-11 17:43:07 UTC; 8s ago
    Process: 4946 ExecStart=/usr/lib/sqm/start-sqm (code=exited, status=0/SUCCESS)
   Main PID: 4946 (code=exited, status=0/SUCCESS)
        CPU: 165ms

root@clearfog:/home/alex# tc qdisc show dev wan
qdisc cake 802f: root refcnt 2 bandwidth 5795Kbit diffserv4 triple-isolate nat nowash no-ack-filter split-gso rtt 100ms ptm overhead 34 mpu 68 
qdisc ingress ffff: parent ffff:fff1 ---------------- 

The problem was before as a dependent of multi-user.target it never retriggers (as multi-user is not restarting). As a dependent of the actual device you want to apply the controls to, when it is (re-)created systemd automatically fires it off.

Fortunately we avoid any possible race as tc of course does not depend on the interface being configured, only that it exists even if it is down.

[tohojo]

Alright, awesome! Thanks for the patch!