dhcpcd not renewing lease after carrier lost

[erg_samowzbudnik]

Hi.
TLDR: On a headless raspberry pi 3 when router goes down and comes back up, wifi connection is lost.

I have two network services:
/etc/init.d/net.wlan0
/etc/init.d/net.enu1u1

with corresponding files in /etc/conf.d:
net.wlan0:

Code: Select all

preup() {
        rfkill unblock all
        return 0
}

modules_wlan0="wpa_supplicant dhcpcd"
config_wlan0="dhcp"

net.enu1u1:

Code: Select all

config_enu1u1="192.168.100.177/24"
dns_servers_enu1u1="192.168.100.1 8.8.8.8"

On startup, everything works fine.
If I restart any of those services, things work fine.

Things get hairy when wireless router goes down and comes back.
Those are the relevant entries from /var/log/messages:

Code: Select all

Feb  5 13:09:53 rpi dhcpcd[1891]: wlan0: carrier lost
Feb  5 13:09:53 rpi wpa_cli[3882]: interface wlan0 DISCONNECTED
Feb  5 13:09:53 rpi dhcpcd[1891]: wlan0: deleting route to 192.168.0.0/24
Feb  5 13:09:53 rpi dhcpcd[1891]: wlan0: deleting default route via 192.168.0.1
Feb  5 13:09:55 rpi dhcpcd[4080]: sending signal TERM to pid 1891
Feb  5 13:09:55 rpi dhcpcd[4080]: waiting for pid 1891 to exit
Feb  5 13:09:55 rpi dhcpcd[1891]: received SIGTERM, stopping
Feb  5 13:09:55 rpi dhcpcd[1891]: wlan0: removing interface
Feb  5 13:10:45 rpi dhcpcd[4080]: pid 1891 failed to exit   # <-- this is dhcpcd failing to exit
Feb  5 13:11:29 rpi wpa_cli[4263]: interface wlan0 CONNECTED
Feb  5 13:11:30 rpi rfkill[4354]: unblock set for all
Feb  5 13:11:31 rpi dhcpcd[4449]: sending commands to dhcpcd process
Feb  5 13:11:31 rpi dhcpcd[1891]: control command: dhcpcd -m 2003 wlan0

So dhcpcd fails to exit and no IP is set on wlan0.

I noticed that restarting net.wlan0 service fails to terminate dhcpcd process just the same.
If I kill dhcpcd and restart the service, things work.
If I stop net.wlan0 service and run wpa_supplicant and dhcpcd manually, network gets back up when router goes down and comes back up.

My /etc/dhcpcd.conf contains default settings.

So clearly the issue is with dhcpcd failing to exit when it is run as a part of a net.wlan0 service.

I'm failing to understand why and what to do about it.

[gentoo_ram]

I would recommend trying to use iwd instead of wpa_supplicant to handle Wifi connections. I've found it to be much more reliable than wpa_supplicant. As a bonus, the daemon is a lot more compact. Follow the wiki directions to configure using iwd. I use it on both of my RPis (4 and 5).

From /etc/conf.d/net:

Code: Select all

modules_phy0="iwd debug"
modules_wlan0="iwd debug"
iwd_wlan0="phy0 debug"
config_wlan0="dhcp"
dhcp_wlan0="nontp"

preup() {

rfkill unblock all

return 0

}

Be sure to install net-wireless/iwd.

Just a couple daemons are created:

Code: Select all

 2744  2744 root      0.0  0.0 Ss   ?        00:00:26  3808  16244 /usr/libexec/iwd -p phy0 -i wlan0
 2791  2790 root      0.0  0.0 S    ?        00:00:11  2060   2784 dhcpcd: wlan0 [ip4] [ip6]

[Zucca]

FWIW, I just switched from networkmanager (because it became unprdictable) to dhcpcd + wpa_supplicant (I also ditched connman on my laptop for these two).
But I don't use the net.* scripts. I configured both manually. Both tools have optional GUIs, but I haven't tested them yet.

[erg_samowzbudnik]

Worked around it by adding to /etc/conf.d/net.wlan0:

Code: Select all

predown() {
killall -6 dhcpcd
return 0
}

Not ideal but it does the job.
Will not mark it fixed as it's a hack, service should be able to terminate dhcpcd without this hack so it's still a mystery to me.